Docstoc

Thesis

Document Sample
Thesis Powered By Docstoc
					SOME APPROACHES TO THE DEVELOPMENT OF A FRAMEWORK
   OF INDICATORS TO MONITOR FINANCIAL STABILITY IN
               DEVELOPING COUNTIRES

       ON THE EXAMPLE OF RUSSIAN FEDERATION
                   MASTER THESIS




                            M.Sc. in Finance and International Business

                                                  Author: Anna Sereda

                                                       Exam #: 277646

                               Academic Advisor: Philipp J.H. Schröder




                    September, 2009
Contents:                                                                                  Page:

Introduction                                                                                (3)

       1.      Scientific background of the problem, description of the financial stability (5)
                   1.1. Banking crises                                                      (8)
                             1.1.1. Theoretical models of banking crisis                    (8)
                             1.1.2. Empirical analysis of banking crisis                    (10)
                   1.2. Currency crises                                                     (15)
                             1.2.1. Theoretical models of banking crisis                    (15)
                             1.2.2. Empirical analysis of banking crisis                    (25)
       2.      Main approaches to the development of a framework of indicators monitoring
       financial stability                                                                  (27)
                   2.1. Qualitative analysis                                                (27)
                   2.2. Econometric evaluation                                              (35)
                   2.3. Nonparametric analysis                                              (42)
       3.      Development of a framework of indicators monitoring financial stability      (47)
                   3.1. Signaling approach                                                  (49)
                   3.2. Analysis of operational capacity of the potential signaling indicators based
                   on the example of Russian Federation in 1994-2009                        (57)
                   3.3. Composition of financial stability indexes                          (60)
       4.      Monitoring the financial stability in Russia 2009 (II quarter)               (65)
                   4.1. Russian Federation (I quarter 2008 – II quarter 2009)               (67)
Conclusions                                                                                 (69)

Literature                                                                                  (71)

Appendix 1-8




                                                                                                   2
Introduction


       Not even the highly developed countries are fully secured from the risks and threats of the
financial system instability. Transitional and developing economies are especially open and
vulnerable to these risks, because while their markets are open, the mechanisms which buffer the
negative impact of the fundamental factors of financial instability are not yet formed. During the
past three decades, the world faced many bouts of financial instability within individual
countries, which sometimes spread world-wide. Recent examples include the sharp price
movements in U.S. equity markets in 1987 (“black Monday”) and 1997; bond market turbulence
in the G-10 countries in 1994 and in the United States in 1996; currency crises in Mexico (1994–
95), Asia (1997), Russia (1998); the collapse of the hedge fund Long-Term Capital Management
in 1998; the currency swings of the 1990s; volatility of global equity markets in 2000 and 2001,
and present full scale world economic crisis are no exceptions to this rule.
       Defining episodes of financial instability is a complicated task. In particular, it is logical
to consider bankruptcy of a few financial institutions as a financial instability. During certain
economic conditions it could be just a part of the regular market process, when unprofitable and
ineffective entities leave the competition. At the same time, bankruptcy of one financial
institution can become a trigger to the financial crisis. Thus, in my work financial instability is
defined as problems in the financial system of the country which cause significant negative
influence on the economic activity.
       Previous scientific research has shown that periods of financial instability prior to
financial crises may have common features. The significant losses that economies face as a result
of the financial crises led to the creation of empirical models that allow recognition of symptoms
prior to the critical point to give policymakers time to neutralize the negative consequences.
Monitoring of the present condition of the financial system based on a series of indicators
analyzed on a regular basis is also important. This fact was realized after significant direct
(recapitalization of the banking system) and indirect (recession) costs of the financial crises
occurred in different regions around the globe in 1990-s.
       Sources of financial instability vary. Among them is a mismanagement of assets, scarcity


                                                                                                   3
of capital in financial institutions, bank runs, etc. When financial instability increases even
insignificant misbalances in economy may lead to the development of the financial crisis
with very negative consequences for the economy.
         In my research I made an attempt to examine possible qualitative, quantitative and non-
parametric characteristics of the present condition of the financial system of Russian Federation.
Analysis of indicators cannot be a strictly formalized procedure because the threshold levels
signaling about decrease/increase of the probability of financial instability are relatively
provisional and depend on the certain economic conditions. It is always necessary to analyze the
present economic situation and in the relation to that correct conclusions drawn on the basis of the
formal analysis.
         The first part of the research defines the relations between particular variables of the
financial stability and examines the scientific expertise. Two of the most common types of crises
are examined in detail – banking and currency crises. Banking crisis is usually related with an
inability of some banks to carry out their liabilities or with the active government interference
focused on the prevention of the occurring problems. Currency crisis is the situation when
speculative attack on the national currency leads to the sharp devaluation which government tries
to prevent by engaging gold and foreign currency reserves or by a significant raise of the interest
rates.
         Notably in the modern history the Russia faced different types of financial crises and
experienced to its fullest their negative consequences. Thus, financial instability by definition is
dangerous and undesirable for sustaining gradual and successful economic development, and
development of the methodology to identify these crises in their early in its early identification is
an important and relevant subject.
         On the basis of conclusions drawn from the first part of work an attempt was made to
establish a framework of indicators to monitor financial stability and separate methodological
matters related to it are examined in the second part of the research. Finally, the working
capacity of the offered framework is tested on the example of the developing market of the
Russian Federation.
         While the application of the offered methodology does not necessarily allow forecasting
with absolute certainty an approaching financial crisis, a framework of signaling indicators with
a certain degree of reliability allows identifying negative tendencies in the economy in advance


                                                                                                   4
and gives an opportunity to neutralize them. Moreover, I will show that methodology of defining
threshold levels means maximization of the prognostic power of the indicators.
CHAPTER 1
Scientific background of the problem, description of the financial stability and its main
components.


       Through the past decades many researchers have attempted to determine a set of indicators
to examine of current tendencies in the development of the financial system. Early surveys
stressed the analysis of the fundamental economic indicators, while modern surveys identify the
investor’s expectation as having the most the important role in forecasting of financial crises.
       Classic interpretation of the financial instability was given by Irving Fisher (Fisher,
1933). He asserts that financial instability is strongly correlated with the macroeconomic
cycles; in particular, with the dynamics of total debt in economy. Problems, related to the over
accumulation of the total debt in the real sector, lead to a situation where it is necessary to
discharge debt in order to restore the equilibrium to the economy. This discharge of debt results
in a decrease of deposits and a divestiture of assets at a low price. This in turn leads towards a
recession with a drop in the rate of price increases and output, as well as an increase in
unemployment and the number of bankruptcies. Thus, according to Fisher, the main cause of the
financial instability is the negative dynamics of the fundamental indexes.
       In their work, Diamond and Dybvig (Diamond, Dybvig, 1983) relate causes of the
financial instability with factors which influence behavior of the banks depositors. They
propose the possibility when the economy transitions from the state of the ‘good’ equilibrium to
the ‘bad’ it can be accompanied by a bank panic. Diamond and Dybvig believe that economic
agents making bank deposits during this period provide some stability of the financial system
and in case of some negative events the probability of bank panic increases. They identify that
the investor’s confidence is an important factor of the financial system stability.
       Mishkin (Mishkin, 1996) examines the role and influence of the asymmetrical
information on the development of the financial system . He asserts that information
asymmetry between creditors and borrowers leads to the creation of the adverse selection. In
other words, borrowers often possess more information about the characteristics of investment
projects in which they intend to invest. Creditors, with incomplete information, are forced to


                                                                                                   5
hedge against risks of uncertainty by lending money under average rate of interest between risk
and zero risk investments. As a result, borrowers needing money for financing high-yield projects
with low risk degree are forced to pay higher interest than they would pay in case of informational
transparency. At the same time, borrowers who finance high risk projects have an opportunity to
obtain loans with lower interest rate. All that leads to displacement of ‘good’ investment projects
with ‘bad’ and therefore to the drop in quality of portfolios of financial intermediaries.
       From the point of view of both creditors and borrowers, Guttentag and Herring
(Guttentag, Herring, 1984) discuss the possibility of occurrence of difference in the expected
return on the project in case of uncertainty on the future return on the investments. For example,
if the expected return on the project for the creditor is lower than return on the alternative project,
the borrower’s loan may be rejected. Guttentag and Herring state that the growth of the financial
instability increases the number of rejections on the loans as well and in its turn this leads to the
instability in the real sector, invoking a new cycle of financial crisis. Deposit insurance is
noted as a possible solution of the problem. However, Keeley (Keeley, 1990) says that deposit
insurance may be related to the problem of moral hazard, and as result financial intermediaries
will face higher risks than in case of not using deposit insurance. The case is that they can
receive money at a risk-free rate (on insured deposits rate) and invest them in high risk projects.
It also can increase sensitivity and vulnerability of the financial system to possible shocks.
       Some studies show that asymmetrical information on the financial markets might be the
source of contagion effect. In the conditions of mutual interdependence between financial
markets of different countries, the negative external shocks might be passed to the wealthy
economies from the others. Kodres and Pritsker (Kodres, Pritsker, 1998) specifically develop a
theoretical model which includes factors affecting the contagion effect, where contagion depends
on the level of information asymmetry. They also demonstrate that even if the mechanisms of
risk hedging exist, the contagion still might happen without influence of negative microeconomic
shocks. That is possible when investors decide to lower risks and offshore financial means out of
the country. In fact, a similar situation took place in Asian countries before the crisis in 1997.
       According to Davis (Davis, 1996) institutional investors often face the agent-principal
problem and it might be one of the facilitators of the financial instability. The case asserts that
the goal of fund managers might not be client’s profit maximization. As a result, the asset price
fluctuations on the market may significantly increase, thereby increasing the probability of


                                                                                                     6
financial instability. To solve this problem the authors suggest monitoring closely top
management activities and apply various evaluation systems to monitor quality of management
performance. In this case, managers will tend to behave like other players on the market and
avoid non-reliance. Following after, other investors will allow top managers to maintain their
reputation on the good level because the risk of achieving results lower than average will
significantly decrease.
       Many researches published during the period of Mexican crisis in 1994 and Asian crisis
in 1997 examined the fragility of financial institutions in relation to exogenous shocks. Authors
pay attention to such factors of financial instability as depreciation of the national currency,
decline of the rate of economic growth, deterioration of the balance of payments, high inflation
rates, deterioration in the terms of trade, speculative attacks on the stock market, and production
loss in the export sectors. Additionally, such quality factors of instability like insufficient
supervision over the banking system, inadequate fiscal and monetary policy, imperfect
legislation, accounting standards and others are being analyzed.
       There is a significant amount of research, the main purpose of which was the detailed
analysis of contagion effect (Baig and Goldfajn, 1999; Fratzscher, 1998). Factors initiating the
contagion included a high correlation between exchange market and stock market, close inter-
country banking and external trade relations, low level of gold and foreign currency reserves, as
well as overall weakness of the financial system. Besides that, Kaminsky and Reinhart
(Kaminsky, Reinhart, 2000) shows that additional risk factor for occurrence of contagion is
the existence of common creditor with the country already experiencing financial crisis.
       Considering that financial and exchange crises often accompany each other, factors
that are used to forecast the exchange crises, might be used as well as the financial crisis
indicators. Thus, correlation between financial and exchange crises was found in the research of
Dornbush, Goldfajn and Valdes (Dornbusch, Goldfajn, Valdes, 1995), Kaminsky and Reinhart
(Kaminsky, Reinhart, 1999), also Kaminsky, Lizondo and Reinhart (Kaminsky, Lizondo,
Reinhart, 1998). Findings show that exchange crisis leads to the financial crisis if the influence of
devaluation on the quality of bank assets is so significant that it decreases significantly net bank
asset value.
       Thus, I reviewed scientific researches on the influence of the macroeconomic indicators
on the stability of the financial system. However, aggregated microeconomic indicators play no


                                                                                                   7
lesser role in the development of a framework of indicators monitoring financial stability. Their
usage in empirical researches is only limited with data accessibility. Thus, in the classic work of
Altman (Altman, 1968) qualitative indicators, such as asset quality, profitability and liquidity are
examined. However, in this case analysis was applied to the individual companies. Later, other
research was conducted which used aggregated microeconomic indicators for the monitoring of
financial stability. Specifically in the works of Frankel, Rose and Honohan (Frankel, Rose,
1996; Honohan, 1997) the importance of such indicators as short-term liabilities in foreign
currencies is noted. Gonzales-Hermosillo, Pazarbasioglu, and Billings (Gonzalez-Hermosillo,
Pazarbasioglu, Billings, 1997) produce testimony to their statement that essential role in
forecasting financial crisis play such factors as non operating loans and capital adequacy.
       In the framework of indicators monitoring financial instability are often included such
factors as interbank rate, relation of deposits volume to money quantity, stock market index and
others. Notably, Kaminsky, Lizondo and Reinhart came to conclusion that aggregated
microeconomic indicators are better to use for forecast of currency crises than banking ones.
       Thus, I reviewed the main scientific studies that examine factors used by researchers as
indicators of financial instability. Now I will examine in detail the banking crises, because
historically they were the type of financial crisis most often to occur.




1.1. Banking crisis


Theoretical models of banking crises


       Specific character of the activities of commercial banks does not allow a full explanation
of their behavior on the basis of standard microeconomic models. Developed theoretical models
and industry analysis of the banking sphere show that while conducting behavioral research of
commercial banks, the problems of asymmetrical information become increasingly important.
These include the relations between “bank – borrower”, “bank – depositor” and credit rationing.
The absence of explicit effects of scale in banking service offers and the specifics of price
competition between commercial banks, creates a limit in the possible application of standard


                                                                                                  8
models in industry analysis. Concurrently, the number of unique market and commercial risks
inherent only to the banking sphere are not discussed in the theory of the firm. Thus, in relation
to the problem of sustainability in a banking firm, the possibility of bankruptcy is one of the
most important aspects of the activity analysis of the commercial banks 1.
         The concept of exposure (vulnerability) of the national banking system was first
introduced in 1977 by Minsky (Minsky, 1977). Traditionally, banking crises are divided into two
main types: bank runs, which spread to a few separate banks including the largest national bank
institutions; and bank panics, when crisis developments spread not only to the entire banking
system but also into the system of national accounts and balances2. The model of Diamond-
Dybving (Diamond, Dybvig, 1983) is considered to be a fundamental model of the banking crisis,
which develops in consequence of liquidity problems of a separate commercial bank.
         This model describes behavior of the commercial bank and its depositors in the
conditions of uncertainty. It is assumed that the bank is attracting deposits which are able to be
withdrawn on demand only and depositors can withdraw their money at any time. Thus, there is
a risk of the possibility that the situation will arise when all depositors would like to withdraw
their money at the same moment of time3. The model allows us to draw three main conclusions:
    1. Banks with attracted demand deposits may stabilize their position on the competitive
         market by sharing risk of deposits withdrawal by depositors, who have different
         intertemporal preferences.
    2. Even though an increase in the amount of demand deposits in liabilities helps to coinsure
         risks of premature deposits withdrawal, it might also lead to undesirable equilibrium,
         when all depositors surrender to panic and desire to withdraw their deposits.
    3. “Bank runs” have serious economic consequences because even “healthy” banks might
         experience problems after premature deposits withdrawal forcing a stop-down of
         investment projects.
         As a possible solution to the problem of premature deposit withdrawal the authors
suggest a system of state insurance of the bank deposits, which will ensure the “good”
1
   Detailed review of microeconomic models describing behavior of the banking organization, also models of
industrial organizations in the banking sectors was presented in (Freixas, Rochet, 1997).
2
  It is necessary to note that bank panic is often a consequence of the bank runs, if the government actions on bailing
out and support of problematic banks were not sufficient enough.
3
   In fact, to experience problems with liquidity, it is not necessary that all depositors address the bank
simultaneously. Because bank assets have different time structure and liquidity, it is enough if the amount of claims
that are higher of the liquid assets of the bank.

                                                                                                                     9
equilibrium; besides its positive effect, such a system has its negative sides as well: state
insurance might lead to the increase of the financing of the highly risky projects, and therefore
will increase the possibility of the bank crisis.
       Further developments of models of “bank panic” and “bank run” are based on the
different modifications of the Diamond-Dybving model:            Chari and Jagannathan (Chari,
Jagannathan, 1988) present a situation when “bank panic” starts as result of
misinterpretation of the real situation by a large number of the depositors ; Temzelides
(Temzelides, 1997) shows that “good” equilibrium can be obtained in the case of even few
banks, though the possibility of bank panic increases in proportion of increase of the average size
of the bank. Chen (Chen, 1999) studied the influence of the contagion effect on the
development of the banking crisis and transformation of “bank run” on the separate banks into
overall “banking panic”.
       Additional focus was placed on the studies of the bank crises from a macro level.
Mishkin, Edwards and Vegh (Mishkin, 1996; Edwards, Vegh, 1997) view problems of separate
banks in the context of the development of the financial crisis. Mishkin specifically pays
attention to the problem of asymmetrical information, its role in the development of crisis
throughout the banking system, and on financial and real sectors of the economy. Edwards and
Vegh show that shock changes of macroeconomic conditions in situations of a predetermined
exchange rate of the national currency might cause the crisis in the banking sphere. This in
turn empowers crisis developments in other sectors of economy.
       For the analysis of the crisis developments in Japan in 1990’s, Hayashi and Prescott
(Hayashi, Prescott, 2002) use the model of the real business cycle and came to the conclusion
that limitations on credit accommodations did not result in a decrease of investment activity.
       With the help of the model of overlapping generations, Barseghyan (Barseghyan, 2004)
studied the government’s role of curing banks during financial crisis and came to the conclusion:
if the government is too slow with curing and rehabilitation procedures, then banks start the
Ponzi game. In other words, take new loans to refinance old ones, which in case of financial
crisis will only worsen the situation.


Empirical analysis of banking crises



                                                                                                 10
       There is a significant amount of scientific research containing empirical analysis of the
crises of national banking systems and bankruptcies of separate banks. Most of them were
prepared by large international financial organizations (World Bank, International Monetary
Fund) which traditionally develop advisory notice and practical measures focused on negotiation
of crisis developments in the banking sphere.
       During the past decades, many of developing and emerging economies faced such
banking crises including, Mexico, Argentina, Thailand, Korea, and Russia. These events were
incredibly destructive not only to the banking sector but to the entire national economy as a
whole. It highlighted the necessity of timely forecasting of the crisis developments so that
necessary measures to prevent them would take place and be effective.
       There are also a significant number of scientific researches into the economic literature
that studies the mechanisms launching the crisis. Most of the research focused on the monitoring
of financial stability is possible to divide in two main groups, depending on the approach used.
       Group one uses “signaling approach” which consists of an observation of indicators
during the “calm” period, the period before the crisis and during the crisis. Then on the base
such analysis a conclusion is drawn about the usefulness and applicability of the certain
indicators for the monitoring financial stability. Significant changes in the dynamics of the
particular indicator before the crisis speak about its applicability for the effective monitoring.
       Group two researchers use the evaluation of econometric models, where as the
endogenous variable used binary variable that is equal to 1 during the crisis or before it. In such
models regressors are different variables used as indicators of financial instability.
       Eichengreen and Rose (Eichengreen, Rose,1998) outlined five main reasons of the crisis
in the banking system. They are: internal macroeconomic policy; external macroeconomic
conditions; exchange rate fluctuations; financial structure of the country; and problems of control
and regulation. They also examined nine groups of variables:
      Variables responsible for the world interaction: volume of international gold and foreign
       exchange reserves (in % from imports volume of the country in a month in monetary
       terms), foreign debt (in % GDP), current account balance of the current account
       transactions (in % GDP) and real exchange rate;
      Key macroeconomic indicators responsible for the fiscal and monetary policy and
       economic dynamics: budgetary deficit/surplus (in % GDP), growth rate of the domestic

                                                                                                     11
       credit and growth rate of GDP per capita in real terms;
      External variables: growth rate of GDP in real terms in the countries of OECD and world
       exchange rate (that is weighted average interest rate of the USA, Germany, Japan,
       France, Great Britain, and Switzerland interest rates; weights were taken proportionally
       to the share in the foreign debt of the examined country, for each of the noted countries).
       Results and outcomes of their research:
           a) World interest rate starts to rise approximately 2 years before the crisis and
               reaches its highest point either a year before the crisis or in the year of crisis. GDP
               growth ratio in real terms drops during the years preceding the crisis, though not
               as much as interest rates.
           b) Other external variables do not have an significant influence.
           c) Such indicators as credit boom and growth of budget deficit cannot be
               considered as signaling indicators of banking crisis because their occurrence
               does not justify about high possibility of crisis any time soon.
       Demirguc-Kunt and Detragiache (Demirguc-Kunt, Detragiache, 1998) also analyzed
variables which could serve as signaling indicators. Given observations were based on the
sample of 31 cases of banking crises from 1980 to 1994 with the usage of logit models. The
authors show that a low growth ratio of the economy, high inflation and high real interest rates
are factors that evidence about rise of possible problems in the banking sector. At the same time,
high growth ration on credits and unfavorable shocks of trade conditions effects low on the
possibility of crisis on banking market. With that the rate of change of the exchange rate and
volume of budget deficit probably do not have significant influence on the possibility of crisis-
producing situation in the banking sectorа. There is an interesting finding concerning the
existence of the deposit insurance system. With introduction of the deposit insurance system in
the country, banks involved in the system start to take higher risks which lead to the elevation of
risks in the banking system overall and logically leads to the increase of possibility of crisis.
       A slightly different approach to identify signaling indicators is used by Hardy and
Pazarbasiogly (Hardy, Pazarbasioglu, 1998). A main variable characterizing the start of banking
crisis, they used a fictitious variable which takes on three values: 2 – during the period of
difficulties in the banking sector, 1 – during preceding period, and 0 – in all other cases.
Authors use the following arguments for defining preceding period before crisis into a separate

                                                                                                    12
category: first, with such separation it becomes possible to determine predictive capability of
signaling indicators without information that is only becomes available during the crisis
period; secondly, behavior of many economic variables might significantly differ during the
crisis and period before it.
          In the sample for the analysis 50 countries were included, 38 of them experienced 43
banking crises varying in severity. In the sample were not included countries which experienced
hyperinflation during the examined period and former socialistic countries with transitional
economies.
          Possible indicators of banking crisis were divided in three main groups: variables of the
real sector; variables of banking sector, and shocks that influence the situation in the banking
sector.
          Besides the noted variables, also taken into account was whether a country belonged to
a particular region and the existence of multiple economic crises in the country.
          Empirical test gave following results:
          a) Banking crises start simultaneously with the significant decrease of GDP growth ratio
             in real terms, sharp rise of consumption alarms about possible crisis in the short run.
          b) Sharp drop of banking deposits in the real terms and significant increase of loans
             issued to the private sector signals about crisis in short run. Considerable increase of
             aggregated foreign liabilities of the country (in % GDP) may also be a signaling
             indicator though its predictive value is lower than in mentioned higher indicators.
          c) First to point on the possibility of crisis is sharp falls of inflation with its following
             growth. Real interest rates usually rise in the period preceding the crisis and keep
             growing during the crisis. Growth of real effective exchange rates, a sharp drop of
             growth ratio on import in real terms also testifies about upcoming banking crisis.
          d) In the countries already experienced in crisis the possibility of its repetition is higher
             than in the countries which have never experienced banking crisis.
          Caprio and Klingebiel (Caprio, Klingebiel, 1996b) came to the conclusion that a poorly
developed financial market does not protect a country from the external shocks.
          Kaminsky (Kaminsky, Reinhart, 1998) study predictive validity of different signaling
indicators. According to their findings the best signaling indicators for forecasting crisis in
banking sphere are from:


                                                                                                    13
      variables characterizing fiscal and monetary policy: relation of M2 to the monetary base
       and relation of domestic credit to GDP;
      variables of current account: export and real exchange course;
      variables characterizing capital markets: difference between real exchange rates (foreign
       and domestic), world interest rate, debt to the banks and deposits dynamics;
      variables characterizing the real sector of economy – dynamics of industrial production,
       internal interest rate in real terms and changes in share indexes.
       Bell and Pain (Bell, Pain, 2000) conducted a brief survey of some previous research on
the matter. Based on these, they draw a conclusion that a banking crisis is usually preceded by
following events: expansion of interest rates in real terms, low growth ratio of output, rapid
growth of domestic credit and drop of volume of external trade and real exchange rate.
       Continuing their previous work (Demirguc-Kunt, Detragiache, 1998), Demirguc-Kunt
and Detragiache conducted an analysis of indicators of banking crises by using newer time-series
data and more countries in the series, in comparison with the previous work. They identified the
following results: low rate of growth of GDP, high inflation and high real interest rates result in
the high probability of the coming banking crisis. Among variables of banking sector, two of
them increase the probability of banking crisis: increase of ration of monetary base to gold and
foreign currency reserves and volume of loans to the private sector. Besides that, authors
showed that low level of GDP per capita and existence of deposit insurance system increases
the possibility of banking crisis in the country.
       Besides that, Demirguc-Kunt and Detragiache systemize possible causes of problems in
banking sector: 1) degree of separate bank expose and systematic banking crises; 2) financial
liberalization and crises; 3) international shocks, regimes of exchange rates and crises; 4)
structure of bank owners and crises; 5) role of institutions, 6) political system and crises.
       In their work Peresetsky, Karminsky and Golovan (Peresetsky, Karminsky, Golovan,
2004) actually show that with the help of the signaling indicators it is possible to predict
problems in the banking system of the country.
       Thus, I reviewed main scientific surveys devoted to the signaling indicator analysis
forecasting banking crisis. Putting them together, we can define following key figures, which
signal about increase of possibility of banking crisis in the best way:
      Economic growth ratio: GDP growth ratio of in real terms, dynamics of industrial

                                                                                                14
       production.
      Balance of payments: eeal effective exchange rate, real exchange rate, import and export.
      Interest rates: real exchange rate, difference between world and domestic interest rates
       in real terms
      Monetary indicators: inflation, GDP deflator, monetary multiplier.
      National income account: consumption.
      Banking sector: bank deposits in real terms (in relation to GDP); loans to the private
       sector in real terms, aggregated foreign liabilities in relation to GDP.




1.2. Currency crisis


Theoretical approaches


       There are three generations of models of currency crises in the economic literature. In the
first generation models, all variables are determined and the depletion of gold and foreign
exchange reserves is considered to be the main cause of currency crises. In the second generation
models, uncertainty plays the most important role; in other words, economic policy is not
predetermined but instead depends on the current state of the economy and economic conditions.
Finally, the third generation models pay close attention to the contagion effect.
       Discussion on the matter of balance of payment crisis exploded after Paul Krugman’s
work (Krugman, 1979). Analogical models become models of the first generation. In his work,
Krugman expresses his hypothesis that the main cause of currency crises is the economic policy,
specifically the inconsistence between effective intervention of fixed exchange rate and
stimulating (fiscal or monetary policy) internal economic policy (monetary expansion).
Krugman describes the situation in which domestic credit growth along with a fixed exchange
rate leads to a reduction of international reserves and, as result, to the boost of speculative
activities with the currency. These activities momentarily exhausted reserves and forced the
government to abandon the policy of fixed rate. Thus, the period before crisis might
accompany a gradual decrease of foreign reserves, budget deficit (fiscal expansion) and a sharp


                                                                                               15
increase in the number of loans to the governmental and private sector (monetary expansion).
        A core model of currency crisis of the first generation is the model of a medium country
where conditions of purchasing power parity and interest rate parity are fulfilled. It is assumed in
the model that economic agents are perfectly anticipated and that domestic assets are available
on the world market and are perfect substitutes to the external assets. While the amount of
international reserves exceeds some minimally accepted value, the central bank controls part of
international reserves to support the fixed exchange rate. When the support of fixed exchange
rate is abandoned due to the speculative attack and reserves drop below minimally accepted
value, the transition to a floating exchange rate takes place. Foreign currencies inside of the
country are not adopted, and all goods are traded on the world market. Thus, the model consists
of five equations:
       Demand equation on money supply:
    m D  p  a 0  a1 y  a 2 i ,                                                  (1)
where, mD – is the logarithm of monetary base, p – is the logarithm of price index inside of the
country; y – real gross national product; i – return on assets in national currency (demand on real
cash balances is due to transactional component and costs of carrying cash).
       Equation of purchasing power parity:
    p = p* + s,                                                                     (2)
where, p* - is the logarithm of external prices index; s – is the logarithm of exchange rate. In the
conditions when all goods are traded and fixed exchange rate is supported, internal prices fully
depend on external prices.
       Equation of interest rate parity:
    i  i *  Es
                                                                                   (3)
where, i* - is external interest rate; Es – expected rate of change of the logarithm of foreign
exchange rate (expected value of derivative of the logarithm of foreign exchange rate in time).
Yield of domestic assets takes in account expected rate of change of the exchange rate of the
national currency. In the conditions of perfect information with the effective intervention of
fixed exchange rate, the expected rate of change of the exchange rare equals zero and domestic
interest rate equals to external interest rate.
       Equation of money supply:
    M S = RS + D,                                                                   (4)
                                                                                                 16
where, MS – monetary base; R – national central bank’s reserves in foreign currency; S –
nominal exchange rate (s=log(S)); D – domestic credit (respective balance sheet accounts of the
central bank).
      Balance one the money market:
   M D = M S,                                                                        (5)
where, M D = exp {mD}.
       If one assumes that domestic credit rises with the constant rate µ, then the reserves of the
national central bank will decrease. Let us suppose that after the speculative attack, the central
bank exhausted its own international reserves and on the floating exchange rate wins over the
exchange market. This value of the exchange rate is called the shadow or equilibrium exchange
rate for the fixed exchange rate at a current time.
       The model shows that shadow exchange rate increases along with the growth of
domestic credit, and that allows the calculation of the moment of devaluation. In the conditions
of perfect information, there cannot be a plummeting of exchange rates (the condition of
arbitrage absence), and that means that speculative attack with the following devaluation will
take place exactly in the moment of time when the shadow exchange rate is equal to the set
value, allowing us to calculate the moment of devaluation.
       The model shows that if in the case of future policy for the central bank and government
is certain, devaluation will have three stages: gradual decline of the reserves, rough attack and
post-crisis period during which the exchange rate will not be fixed on the same level. If agents
foresee that attack on reserves might lead to devaluation and abolishment of the fixed exchange
rate, they will act to lower the level of reserves to the minimal level, thus denying for the central
bank an opportunity to protect overvalued currency.
       In the real situation, policies of the government and central bank could not be known
for sure, and that is why the assumption about perfect information is not correct. For the
analysis of the currency crisis under uncertainty, stochastic variables are added in the model
described above. There are two ways of inserting such variables in the model: 1) it is unknown in
advance level of the international reserves; if reserves drop lower than this level, then the central
bank gives up the support of the fixed exchange rate and devaluation of the national currency
happens; 2) is the uncertainty in the dynamics of the domestic credit.
       Uncertainty of the reserve’s volume which the central bank will use for the national

                                                                                                  17
currency protections is introduced in Krugman’s article (Krugman, 1979). The main result of the
model is a finding that speculative behavior is relatively sensitive to the specification of the
process, which sets minimally allowed variables for the central bank volume of international
reserves. If this level is random, then devaluation starts as the result of continuing during
certain amount of time speculative attacks. If the volume of minimal allowed reserves is fixed
and unknown to the currency profiteers, then like in the classic model, devaluations starts as the
result of the sudden speculative attack (Willman, 1989).
       Uncertainty in the dynamics of the domestic credit was first introduced in the article of
Flood and Garber (Flood, Garber, 1984), in the stochastic model in discrete time. If to add
stochastic shock (on which the shadow exchange rate depends) to the constant rate of increase of
the domestic credit, then the shadow exchange rate will be a stochastic variable, that means it is
possible to evaluate the possibility of devaluation as the possibility of the shadow exchange rate
will exceed value set by the central bank in the next period:
          Pr( ~  s )
         t 1    s   t 1                                                              (6)

where, πt+1 – possibility of devaluation in the next time period; ~t 1 – shadow exchange rate in
                                                                  s

the next period; s – set by the central bank value of the exchange course. While making a
decision on the speculative attack, economic agents look up to expected value of the exchange
rate which reflects expected profit as the result of the attack:
        Es  (1   ) s   ~
           t 1             t 1s   t 1 t 1                                          (7)

where, Es t 1 – expected profit of the profiteer in the next period; πt+1 – possibility of devaluation

during the next period;            ~ – shadow exchange rate in the next period; s – set up by the
                                   st 1

central bank value of the exchange rate. Extensions of the crisis models for the slow price
adjustments and the case when domestic and external assets are not perfect substitutes, are
presented by in some research (for example, Flood, Hodrick, 1986; Willman, 1988). The easiest
method of adding of slow price adjustments into the model is the substitution of the equation
of price parity on dynamic equation for process in the form offered by Dornbusch
(Dornbusch, 1976; Dornbusch, 1987):
        pt    [ ( st  pt )    (it  pt )  y ],
                                                                                      (8)

where, ƛ – is the speed of price adjustments in relation to the excessive demand ; st – is the
logarithm of the exchange rate; pt – logarithm of the price index; it – logarithm of the interest

                                                                                                    18
rate; y – aggregated demand, which negatively depends on the real interest rate and real

exchange rate.
       Further, solving system for the floating exchange rate, we find path of the shadow
exchange rate, which allows us to calculate similarly the moment of devaluation. The model
shows that the faster the prices change, the faster the currency crisis starts (T /   0) . If the
price change is momentary, then they remain constant during the period of the support of the
fixed exchange rate and change by jump during the devaluation. If the prices adjust slowly and
an expectation of devaluation takes place, then internal prices will change gradually from the
moment of expectations occur and to the moment of devaluation will achieve its new value. The
slower prices change, the more time is needed to achieve a new value and accordingly the
expected time before devaluation is larger.
       The other method of weakening the purchasing power parity is the separation of the
goods on traded and non-traded, which is offered by Goldberg (Goldberg, 1988; Goldberg,
1994). Traded goods satisfy the parity prices equation, and with the fixed exchange rate good’s
prices change along with the external prices. Prices on the non-traded goods which do not enter
the world market include systematic and random deviations from the parity:
        Pt / S t  Pt *   t   t ,                                                (9)

here, Pt – domestic index on non-traded goods; St – nominal exchange rate; Pt * – external

prices index;  t – systematic price deviations from the purchasing power parity; Ωt – random

shocks. Accordingly, aggregated price index is a weighted sum of price indexes on traded and
non-traded goods.
       Besides that, the model separates random fluctuations of the domestic credit on two
components – unexpected income and unforeseen expenses, random changes of limitations on
accessibility of foreign lending.
       Just like in the basic stochastic model (Flood, Garber, 1984), the main result of this
model is the calculation of the shadow exchange rate path and devaluation probability.
       Blanco and Garber (Blanco, Garber, 1986) modify the currency crisis model of the first
generation for the description for the currency crises in Mexico, with accounting the fact that
devaluation didn’t lead to the transition to the floating exchange rate; instead the new exchange
rate was fixed on a new level. In the model, the main factor leading to the crisis is the fiscal and


                                                                                                  19
monetary policy that leads to the excessive build-up of the domestic credit.
       Fundamental variables of the model cause misbalance of demand on real cash balances,
and it is a stochastic factor which determines random development of the events which
influences the exchange market and described by the equations (1)–(5):
        ~
        ht  ~t    ( E~t 1  ~t ),
             s           s       s                                                  (10)
where, st – is the logarithm of the nominal exchange rate.
       If one assumes, excessive demand is a stochastic process (authors suggest that ht –
autoregressive process of the first order), then the shadow exchange rate which is determined
through ht will also be a stochastic variable. That is why in this model is it possible to talk
about possibility of devaluation and expected exchange rate.
       Assuming that during the devaluation the set value of the exchange rate varies
proportionally with the excess of the shadow exchange rate over the fixed exchange rate before
devaluation (the coefficient of proportionality is constant for each devaluation). This permits the
calculation of the minimum permissible volume of the central bank’s reserves.
       In the work of Cumby and van Wijnbergen (Cumby, van Wijnbergen, 1989), an
analogical model is used for the analysis of the stabilization program in Argentina in 1978–1981.
The specialty of the situation in Argentina was in its stabilization program, which was based on
the tables of the daily values of the exchange rate which were announced in advance. From the
point of view of the model, the announced values can be seen as a fixed exchange rate. If for its
support it required heavily expanding domestic credit, then that might increase the pressure on
the foxed exchange rate, even with the account of its daily fluctuations. The suggested model is
an analogy to the Blanco and Garber’s model with the only difference being that in their research
the authors do not aggregate variables which influence demand on the real cash balances and
exchange market into one variable ht.
       Calculation results on the offered model showed that sharp increases of the domestic
credit in the second quarter of 1980 led to downfall of creditability to the capacity of the central
bank to retain announced exchange rate. With that just prior to devaluation in June, 1981 abrupt
jump of the devaluation probability of 80% in the next period took place.
       For the analytical purposes of the EU exchange rate system during 90’s, a wide range of
models were used, which would examine dynamics of the exchange rate within some interval of
values, exit out of which prevented through money market intervention. The fundamental result
                                                                                                 20
of this research is the conclusion that with the sufficient amount of reserves the exchange rate
might be held within the set limits, with that the path near the target range depends mainly from
the interventions of the central bank, not from the behavior of the fundamental variables
(Krugman, 1991).
       Sterilization of currency interventions is one of the tolerable approaches used by the
central bank for its monetary policy so that drastic reduction of the monetary base during the
speculative attack would not happen. That means additional enlargement of the domestic credit
on the amount equals to the size of intervention in the national currency. Thus, under the
condition of sterilization of currency interventions, shadow exchange rate turns out to be always
higher than the fixed one (Flood, Marion, 1998). That means that in the conditions of full
certainty in the model the implementation of sterilization of currency interventions is
incompatible with the support of fixed exchange rate.
       In real life, we have countries with fixed exchange rates, and simultaneously with the
speculative attack, sterilization of the currency intervention takes place, moreover, such regimes
can sustain for quite a long time. For the account of this event, Flood, Garber and Kramer
(Flood, Garber, Kramer, 1996) express the hypothesis that domestic and foreign assets are not
perfect substitutes, and domestic interest rate includes risk premium for investing into the
domestic assets, which depends on the share of national assets in the investor’s portfolio inside
of the country and abroad. This assumption allows a weakening of the model for the case when
domestic and foreign securities are not perfect substitutes (there are barriers, risk premium for
the investments inside of the country takes place and it means compensation liability related with
the partial substitution of domestic and external securities).
       Inclusion of another variable which reflects the risk premium into the equation of interest
rate parity brings nonlinearity into the model through the behavior of economic agents (private
sector), which might lead to the existence of multiple equilibriums. That means that in the classic
model besides the crisis, which happens as result of inconsistent fiscal and monetary policy,
there is a possibility of situation when crisis happens because of self-fulfilling expectations
(“bad” equilibrium) and possible sudden change from the one equilibrium to another one, which
is more favorable for speculative attack. Research shows that a similar approach entertains the
possibility of existence of the regime with the fixed exchange rate and sterilization of currency
intervention, but this is possible only when reserves are relatively large.


                                                                                                21
        There is one more method of introduction of nonlinearity to the model – assumption
about different regimes of the government actions, when after devaluation the fixed exchange
course is supported (Obstfeld, 1996). Suppose that in result of devaluation government changes
build-up rate of the domestic credit for 1   (build-up rate of the domestic credit before
devaluation). In that case it is possible to calculate two paths of the shadow exchange rate,
accordingly matched µ and µ1. Obviously, higher build-up rate of the domestic credit leads to
the higher pressure on the exchange rate and advances the moment of devaluation, because it
corresponds with the higher value of the shadow exchange rate.
        If the value of the domestic credit is not too large, then when fixed exchange rates are
supported, it is in equilibrium. If domestic credit exceeds some limit, then in the situation of
perfect information, the equilibrium is the speculative attack and devaluation. At the same time,
with the certain values of the domestic credit, the possibility of devaluation depends on level of
coordination of the profiteers (size of the international reserves of the central bank is positive).
If profiteers are disaggregated, then equilibrium will be a preservation of the fixed exchange
rate regime; if they unite, then after the speculative attack devaluation will start. In other words,
in such situation multiple equilibriums are possible; moreover, transition from one equilibrium to
another might happen independently from actions of government and central bank. This
approach is based on the main idea of the currency crisis models of the second generation – there
are different variants are possible of the government’s policy depending on the economy
conditions, profiteers behavior and overall situation on the currency market.
    2   In his article Obstfeld (Obstfeld, 1994) examines the choice between devaluation and its
consequences (influence on unemployment, real output, etc.) as the optimality criterion of
implemented. Let following function of the social costs minimized:
              2       (  E  u  k ) 2
        L                                  min,                                   (11)
              2                 2
where,  – devaluation rate; u – random fluctuations with the zero expected value and
predetermined dispersion; k – measure of devaluation influence (for example, macro
indicators rate of change).
        With the certain correlation between these parameters, it might happen that devaluation
will have to be implemented even when having an opportunity to protect fixed exchange rate with
the help of currency interventions. With that comparison of theoretical values of the parameters,

                                                                                                  22
which are relevant for devaluation, with the actual implementations of the currency crises in
different countries allows us to evaluate costs related to the transition from the one equilibrium
to another.
       In general, versus from the models of the first generation, where exhaustion of the
international reserves is the main cause of the currency crisis, models of the second generation
are based on the hypothesis that government might abandon the fixed exchange rate regime for
the sake of negative influence of such regime on other main economic variables.
       Multiple goals complicate the choice between the support of the fixed exchange rate and
alternative government policies (for example, decline in unemployment, banking system support,
etc.). Target functions for the government might be set as positively dependent from the support
of the fixed exchange rate, and negatively dependent - from deviation of output from the target
level. Under certain conditions costs of the support of the fixed exchange rate might exceed the
benefits. For example, external shock in a shape of increase of world interest rate with the fixed
exchange rate will cause increase of domestic interest rates and drop in output and employment,
leading to the increase of government costs. As soon as world interest rate exceed the certain
level, it is no longer profitable to sustain the fixed exchange rate regime for the government. In a
similar way the other factors influencing the target function of the government might serve as the
potential indicators of the oncoming crisis.
       Models of currency crises of the second generation also show that because of the multiple
equilibrium on the exchange markets, which is related to the nominal character of the
macroeconomic policy, crises might be self-fulfilling with the crises development occurring
without noticeable changes in the fundamental variables. Key assumption in these models is the
fact that macroeconomic policy is not predetermined but transforms depending on the changes
in the economy and economic agents form their expectations taking in account this correlation.
In its turn, expectation and actions of economic agents influence economic variables, thus
affecting overall economic policy. Such cyclicality leads to the potential multiple equilibrium
and economy might transfer from the one equilibrium to another without feasible changes in
the fundamental variables. Thus, initially the economy might stay in equilibrium compatible
with the fixed exchange rate regime, but sudden deterioration of expectations might lead to the
change in policy, and as result, to the abandonment of the fixed exchange rate regime,
confirming by this expectations of the economic agents.


                                                                                                 23
       In the view of the above conclusions, the list of possible signaling indicators might be
complemented with the following variables: output deviation from the defined trend or desired
level; high unemployment rate; growth of world and domestic interest rates; volume of the
national debt; problems in the banking sector; and political variables.
       There is a certain difficulty in the models of the second generation in the explanation of
the financial crisis only on the base on adverse moves of the fundamental variables. Crisis might
develop without substantial modification of the fundamental variables. As result, crisis
forecasting becomes extremely difficult.
       In the works devoted to the models of third generation, attention is paid to the “contagion
effect”, i.e. effect of crisis transfer. In their work Gerlach and Smets (Gerlach, Smets, 1994)
present a model in which devaluation in one country leads to devaluation in the countries –
trade partners because of the desire of the last to avoid loss of competitiveness.
       According to Tomczynska (Tomczynska, 2000) it is possible to determine three factors
which make a separate county vulnerable in case of financial instability in the world. First,
contagion effect with the external financial crises happens when there are system-based relations
between a country and an economy experiencing difficulties. These system-based relations might
be non-observable and, as the rule, based on the similar economic and institutional
characteristics. Second factor that increases the risk of the crisis transfer is the macroeconomic
relations (Eichengreen, Rose, Wyplosz, 1996). To characterize, the last factor is possible through
the definition of a few channels of contagion that affects transfer:
   1) world shocks lead to rise of the pressure simultaneously on the currencies of different
       countries;
   2) significant devaluation of the currency in one country suppresses export of trading
       partners as the consequence of loss of the price competitiveness;
   3) existence of financial relations leads to the situation when crisis events in one encourage
       investors to balance their portfolios through the risk management.
       There is a third factor which might play the key role in the definition of the adverse effect
of the international financial problems. It is reflected in such called “effects of full contagion”.
Full contagion happens when majority of international investors act similar independently from
the condition of fundamental variables. When crisis happens in some developing country, they
can divest assets simultaneously from the developing markets to developed ones (this occurred


                                                                                                 24
during the crises in 1997 in the developing countries).
Empirical analysis of currency crises


       Large numbers of the financial crises in the developing countries in during 90’s kindled
significant interest in the models of early signaling and led to heated discussion on the subject in
the economic literature.
       Thus, extensive review was made by Kaminsky, Lizondo, and Reinhart (Kaminsky,
Lizondo, Reinhart, 1998). The authors examined 28 works devoted to the empirical research of
the currency crises for last 20 years. In spite of the fact that research data significantly differ
from methodological point of view and from the list of the examined crises, the authors drew
some common conclusions. First of all, crises can have many causes, part of which is described
by the dynamics of the separate economic variable. As result, for the explanation of all crises a
large amount of variables might be needed. At the same time dynamics of some variables seem
to have good forecasting characteristics for the prediction of the majority of crises. In
particular, real exchange rate and international reserves were examined in many of the
researches and turn out to be meaningful in most of the cases.
       After publication of the Kaminsky, Lizondo, Reinhart research, all works on the subject
of forecasting financial crises started to be divided in four groups depending on the
methodology used in the research. First group usually includes early works that only includes
the qualitative discussion of causes and events preceding currency crises.
       The second group of works studies distinctive principles in the dynamics of
macroeconomic variables for the period before and after currency crises.
       Non-parametric approach is used in the third group of works and it is based on the system
“signal extraction”. Based on this approach, a conclusion is made about the applicability of
different variables as the signals about approaching crisis.
       The fourth group of the articles estimates the possibility of the financial crisis based on the
explicit theoretical model. In these articles the indicator of the currency crisis is modeled as a
fictitious variable, taking values 0 or 1. Nevertheless, other than in the previous approach,
independent variables do not take a form of fictitious variables, with that regression of all
variables is analyzed simultaneously, while signaling approach studies interrelation of dependent
variable and independent variable are studied separately for each of the variables (Frankel and


                                                                                                   25
Rose, 1996).
        Frankel and Rose used the stochastic model for the crisis probability evaluation on the
annual data from the sample of 105 developing countries from the period of 1971–1992.
Devaluation of the national currency on more than 25% in a year was used for the definition of
the crisis in a country.
        The authors used a few specifications of the model and came to the conclusion that
possibility of the crisis increases when output rate of growth drops down, increases rate of
growth of the domestic credit, international interest rates rise, foreign direct investments
decrease as a part of aggregated debt, international reserves decrease and national currency
strengthens. At the same time results for the output rate of growth, real exchange rate and
reserves were not sustainable in different specifications.
        Berg and Pattillo overestimated the results of Frankel and Rose’s research. Using as an
approaching crisis signal a fictitious variable equal 1, and if possibility of the crisis exceeded
25%, only 17 out of 69 crises were predicted correctly while 33 out of 711 normal periods were
faulty predicted. According to the authors one of the reasons of such results was high diversity of
the sample countries. However, examining more homogeneous and small sample of the countries
for the period of 1970–1996, 38 out of 60 crises and 342 out of 383 normal periods were
forecasted correctly.
        The authors also used stochastic model for the currency crisis probability evaluation.
They use data and the definition of the crisis which is described in the research by Kaminsky,
Lizondo, and Reinhart. In their regression model, the value of dependent variable equals to 1 not
only during the crises but during the periods of 23 months preceding the crisis. Following
variables showed highest predictive value:
       real exchange rate deviation from the determined trend;
       current account position of the balance of payments;
       rate of growth of international reserves;
       export growth ratio;
       relation of the money supply (M2) to reserves.
        Using as an approaching crisis signal a fictitious variable equals 1, and if possibility of the
crisis exceeded 25%, model predicted correctly 48% of the crises inside of the sample and 84%
of normal periods. Out of sample analysis results turn out to be even better: 80% of crises and

                                                                                                    26
79% of normal periods were forecasted correctly.
CHAPTER 2


Main approaches to the development of a framework of indicators to monitor financial
stability.


        All research can be divided in three groups depending on the methodology used to
determine the most effective signaling indicators of the financial crisis.
        Quality analysis. This approach includes a comparison of graphs of the dynamics of the
fundamental economic variables during the period of crisis and at a normal time. To some extent,
calculation of some statistic indicators is available; these indicators characterize dynamics of the
time series of signaling indicators.
        Econometric models. Such an approach implies to the regression models, which allows
evaluation of the interrelation of variables with the financial crisis probability. Most often logit-
or probit-analysis is used, where regression model shows dependence of the financial crisis
probability from the number of economic indicators. Estimated model is used for forecasting
purposes to predict probability of the financial crisis in the future.
        Non-parametric evaluation. Different numerical characteristics are established to define
economic vulnerability tendencies in advance. There are two main directions: 1) determination of
the threshold level for the indicators; 2) establishment of consolidated indexes of financial
stability.
        In literature review, I will present main scientific papers related to each of the noted
groups, though often authors combined several approaches in their researches.




2.1. Quality analysis


        In their work, Eichengreen, Rose and Wyplosz (Eichengreen, Rose, Wyplosz, 1995) make
an attempt to determine variables which can serve as early warning signs of impending crises .


                                                                                                  27
The authors tried to find an answer to the question of whether dynamics of some economic
variables can predict exchange rate crises, they also tried to analyze if the behavior exhibits
different indicators during the period prior to the crisis and after the crisis.
        For the purposes of study of currency crisis the list was made of official announcements
about devaluation and revaluation, transition from the fixed exchange rate to the floating
exchange rate, when thresholds of exchange rate band were broadened, and other significant
changes in the fiscal policies around the world during the period of 1959−1993. Such cases are
called “events” of the exchange market because it is clear that not all of these episodes were full
scale currency crises. Data of twenty different countries was used on a quarterly basis
        Authors presented behavior of the variables during different crises and “events” on the
exchange market in the form of diagrams, hoping to find some common trends and patterns in
their behavior. Each graph illustrated behavior of some variable two years before and after the
“event” or a crisis. They studied dynamics of the following variables: changes in gold and
foreign currency reserves; exchange rate; short-term interest rate; refinancing rate of the Central
Bank; changes in export and import; relation of current account balance of balance of payments
to GDP; relation of budget deficit to GDP; domestic credit; money supply; unemployment;
inflation; GDP in real terms; yield of government stocks and bonds; and stock index.
        Authors show that a few quarters before devaluation decrease of the gold and foreign
currency reserves is observed, export decrease and import increase, and therefore, increase of the
current account deficit. After devaluation reserves and exports recover relatively fast. At the
same time for the import and current account balance more time is needed. Moreover, a
significant increase of the real effective exchange rate before devaluation was discovered.
        In countries before devaluation large budget deficit, growth of domestic credit and money
supply were registered. Higher rates of inflations and short-term interest rates were observed,
notably interest rates did not decrease after devaluation. On the labor market before and after
devaluation high unemployment rates were observed.
        Authors also noted that before revaluation macroeconomic variables show opposite
tendencies, their behaviors do not differ much from their normal state.
        Frankel and Rose (Frankel and Rose, 1996) also used graph analysis to determine
potential signaling indicators of the currency crisis. For their analysis they used data from 105
countries for the period of 1971-1992. In their work currency crisis is defined as nominal


                                                                                                28
devaluation of the exchange rate on more than 25% with the increase of the devaluation speed on
more than 10%. It is noted that even though changes of foreign reserves and interest rates
should be taken in account for the identification of a crisis, it is not done in their work because
necessary data is not available from many of developing countries.
       In the work variables are divided in following groups:
      Domestic variables: growth ratio of the domestic credit; relation of surplus/deficit of the
       budget to GDP; GDP growth ratio in real terms.
      Foreign variables: relation of foreign debt to GDP; relation of gold and foreign currency
       reserves to volume of import, on a monthly basis; relation of current account balance to
       GDP; real exchange rate.
      Variables characterizing foreign debt (in relation to total sum of foreign debt): share of
       foreign borrowing attracted by commercial banks; share of easy loans; share of foreign
       debt with the floating interest rate; share of foreign debt of the government sector; share
       of short-term foreign debt; share of the debt to international financial organizations; and
       the relation of inflow of foreign direct investments to foreign debt.
      External variables: short-term world interest rate       (average weighted interest rate in
       industrial countries); GDP growth ratio in real terms in the countries of OECD.
       Authors show that foreign interest rates rise before the crisis. At the same time account
balance of the current operations of the balance of payments and state budget deficit sustain their
long-term average values before the crisis.
       Eichengreen and Rose (Eichengreen, Rose, 1998) studied banking crises in the
developing countries. They used macroeconomic and financial indexes for the period of
1975−1992. Indicators were divided in five main groups: indicators of the domestic
macroeconomic policy, indicators of the external economic conditions, exchange rate regime,
indicators of the domestic financial structure, problems of supervision and management.
       For identification of the banking crisis authors used the results of Caprio and Klingebeil
(Caprio, Klingebiel, 1996). Banking crisis is defined as a situation when problems that banks
are experiencing lead to the significant decrease of the capital in the banking system. Caprio
and Klingbiel analyzed crisis developments in 69 countries relying on official figures and expert
judgments and made an attempt to determined how serious the problems are and how they
influence the banking sectorр.

                                                                                                29
        Eichengreen and Rose examine dynamics of the nine key variables prior to 39 crisis
events: The authors show that prior to a banking crisis foreign interest rates rise and industrial
output slows down. Though, dynamics of the “foreign” and domestic variables do not allow
using them as signaling indicator of the financial instability.
        In the research of the International Monetary Fund (International Monetary Fund, 1998),
in the identification of the currency crisis, the index is composed equal to weighted average
values of the rate of increase of the exchange rate and gold and foreign currency reserves. They
define a banking crisis as a situation of forced large-scale shut down and mergers of banks or
acquisition of financial institutions by the government; also as massive withdrawal of deposits by
the population. Research examines crises in 50 countries during the period of 1975-1997 and
showed that developing countries experienced larger amount of crises during the noted period. It
turned out that banking and currency crises are often follow one after another. Following
variables were used as signaling indicators: surplus of balance of trade; real effective exchange rate;
gold and foreign currency reserves; export; GDP in real terms; index of stock market; inflation; dynamics
of the money supply in nominal and real terms; relation of the money supply to gold and foreign currency
reserves; relation of the money indexes M2 to M1.
        The following variables were used for forecast of the banking crisis: GDP in real terms;
real fixed capital formation/capital investments; surplus/deficit of budget; inflation; nominal
exchange rate; real effective exchange rate; domestic credit; оrelation the money indexes M2 to
M1.in real terms; index of stock market; the relation of account balance of the current operations
to GDP; and the capital inflow/outflow.
        Examination of the indicators allowed determination of the increase of inflation and
exchange rate, and also a decrease of the export volumes before the crisis. Moreover, for the
period prior to the crisis is typically the increase of the rate of growth of the money supply and
enlarging of the domestic credit.
        Before the banking crisis, the overall situation in the economy is characterized by high
inflation, significant deficit of the current operations account, rapid growth of the domestic
credit, related to some degree to the foreign capital inflow and in some cases with previous
liberalization of the financial system of the country.
        Due to the fact that before the crisis values of not all variables differed statistically
significant from the normal level and due to the availability of the statistical data, as best


                                                                                                      30
signaling indicators of the financial crisis were selected real exchange rate, domestic credit, and
relation of the money supply to gold and foreign currency reserves.
        Glick and Moreno (Glick, Moreno, 1999) study crises in the countries of the East Asia
and Latin America during the period of 1972–1997. They describe crisis as a situation when
deviation of the exchange rate from the mid-value exceeds two standard deviations calculated on
the entire period selected. Periods of hyperinflations are examined separately.
        Dynamics of the following variables is examined in the article:
       Indicators of the currency market: nominal and real money supply M2; monetary
        multiplier; relation of money supply M2 to gold and foreign currency reserves; and
        domestic credit in real terms.
       Indicators of competitiveness and trade: regression of the real effective exchange rate
        from trend (i.e. remains of the regression equation of the real effective exchange rate on
        trend, export, import and relation of net exports to aggregated export); export dynamics;
        and the trade balance surplus.
        Graph analysis shows that increases of the money supply in real terms and domestic
credit in real terms slow down before the crisis, and it speaks about the decline in economic
activity. Such results contradict other similar research, in which sharp increases of the money
supply signals about approaching crisis. Monetary multiplier increases before the crisis and it
becomes obvious approximately one month before the crisis. Authors also found a tendency of
declining gold and foreign currency reserves before the crisis.
        Increase of money supply in nominal terms and growth of the domestic credit in
comparison to the “calm” period were higher in Latin America and lower in Asia, which is
explained with high inflation in the first region.
        Prior to crisis in both regions real export and trade balance surplus were also lower and
real exchange rate was higher than during “calm” periods. Obviously, such dynamics showed
high degree of the exposure to external shocks.
        Aziz, Caramazza, and Salgado (Aziz, Caramazza, Salgado, 2000) study the relations
between macroeconomic and financial variables. One example is between some of the industrial
and developing economies. They studied crises which were revealed in depreciation of the
national currency and sharp decrease of the volume of the gold and foreign currency reserves.
Analysis included comparison of the dynamics of different macroeconomic and financial

                                                                                                31
variables during the crisis and “calm” period.
        The article studied currency crises during the period of 1975−1997. Authors define
currency crisis as significant devaluation of the national currency. A series of indicators whose
behavior significantly differ during the critical events were idetified on the basis of the
conclusions from the theoretical models of currency crises and results from earlier researches on
the same subject.
        For each of the variables whose period of observation (from 1975 to 1997) was divided
between “calm” periods and so called “critical windows”. “Critical windows” are a number of
periods prior to and after the critical date. Authors used windows in 49 months for the monthly
data (24 months prior to the crisis and 24 months after it) and window in 5 years for the annual
data (2,5 years prior to the crisis and 2,5 after it).
        After average values were calculated (on the critical events) for all variables for each
moment of time within the “critical window”. Average values were calculated of the variables
during the “calm periods”. To define whether variable’s behavior differs significantly during the
crisis period from the dynamics of the “calm period”, authors performed standard t-test on the
statistical significance of the deviation in average values of the variable during critical and
“calm” periods. Entire selection of crises was divided in following (possibly crossing) subgroups:
    1. crises in industrial countries;
    2. crises in the countries with developing economy;
    3. crises of devaluation of the national currency (i.e. crises when 75% of increase of
        speculative pressure index is defined by the exchange rate);
    4. crises of significant loss of the gold and foreign currency reserves (i.e. crises when 75%
        of increase of speculative pressure index is defined by loss of the gold and foreign
        currency reserves);
    5. “rough” crises, when speculative pressure index value exceeds three standard deviations from
        the average value;
    6. “soft” crises, when index value is in the interval from 1,5 to 2 standard deviations from the
        average value;
    7. crises cause by the problems in banking sector;
    8. crises with fast following recovery of the economy, when GDP returns to its normal
        trend within two years;


                                                                                                 32
    9. crises with the slow following recovery of the economy, when GDP returns to its normal
        trend within three years or later.
        For all types of crises real exchange rate in average was higher during the “calm” period
in comparison with the crisis period. Two years before the crisis its average value was 0,4
standard deviation higher than during the crisis. In most of the cases after crisis real exchange
rate declined fast.
        Sometimes, currency crises were preceded by a slowdown in the growth rate of export,
even though, it wasn’t significant and was not observed in all types of crises. In particular, it
wasn’t statistically significant for the crises in industrial countries and crises of significant loss
of the gold and foreign currency reserves.
        Terms of trade (relation of export prices to import prices) usually worsens during the
period prior to the crisis, yet statistically significant if it was only few months before the crisis.
Besides, majority of the types of crises had deficit of the balance of payments before the crisis.
        For most of the types of crises inflation significantly exceeded its value of the “calm”
period before the crises, though analysis of “rough” crises and crises related with problems in
banking sphere didn’t reveal such tendency.
        For the study purposes on the monetary market indexes M1 and M2 were used. Analysis
showed increase of the nominal value of the indexes a year – year and a half before the crisis.
Nominal growth of the domestic credit was also noted but it wasn’t expressed strongly. At the
same time growth acceleration of the domestic credit in real terms was not significant. In real
terms indexes M1 and M2 raise approximately 24−12 month before the crisis and after that
(precisely before the crisis) demonstrated decline.
        In most of the cases, a slowdown of the growth ratio of output was not found, though
with significant currency devaluation and crises followed by slow recovery a statistically
significant decrease of the rate of growth of output was still noticed. Finally, an increase of
the world interest rates was registered approximately a half a year before the crisis.
        In their research Carramazza, Ricci and Salgado (Carramazza, Ricci, Salgado, 2000)
used a similar approach in defining critical period. Four different groups of variables are
examined in the article:
    1) General shocks include an increase of the world interest rate, slowdown of the world
        economic growth, decrease of prices on export goods, sharp change of the exchange rates


                                                                                                     33
       of the currencies of the main world economies.
   2) External economic relations. If a country experiences financial crisis accompanied with
       strong currency depreciation, other countries may also get hurt because of the export
       decline in this country and increase of the price competition from the goods imported
       from such country. Thus, necessity of examination of not only the countries which suffer
       from the crisis but also those who can get hurt as well is outlined.
   3) Capital outflow also may be a cause of the crisis or intensify it. Thus, financial instability
       in the country may stimulate investors to change the structure of the investment portfolio
       for the risk cushioning, in other words, cut investments in the assets in this country.
       Obviously, significant capital outflow may increase financial exposure of the country.
   4) Changes in the investors’ expectations can also play a key role in the spread around of
       the crisis. Crisis in one country may serve as a signal that in other countries
       macroeconomic indicators can get worse. With that investors desire to obtain profits
       through the attacks on the currencies of those countries whose situation is similar with
       the one experienced the crisis.
       In most of the developing economies, opposed to developed countries, increase of the
real exchange rate within three years period before the crisis was significant. Balance of
payments deficit a year before the crisis in most cases turned out to be significantly higher than
the average one during the “calm” period.
       Authors also showed that factors which can reflect a country’s exposure to the crises
include slowdown of the GDP rate of growth and high unemployment level. Decrease of output
before the crisis was noticeable mainly in the developed countries. Variables of the monetary
market and budget deficit didn’t enter in the number of variables which allow predicting
developments of the critical events.
       Besides, the relation of the monetary index M2 to gold and foreign currency reserves was
recognized as a working signaling indicator of the financial crisis. Thus, in Mexico this indicator
significantly exceeded its own normal average value during the “calm” period.




                                                                                                 34
2.2. Econometric valuation


       Eichengreen, Rose and Wyplosz (Eichengreen, Rose, Wyplosz, 1995) used quarterly data
from 20 countries of OECD from the period of 1959–1993 to determine the best signaling
indicators of the financial crisis. The authors showed particular interest in political variables. In
their work they showed that control over the capital flows can give the government an
opportunity to repulse a speculative attack on the nation’s currency.
       With further help of logit-analysis, the authors defined macroeconomic variables which
could be used to predict financial crisis. It turned out that speculative attacks as well as transition
to the fixed exchange rate statistically significant increase probability of the financial crisis.
Moreover, in their research, the role of inflation and monetary factors was confirmed as good
signaling indicators of the financial crisis. Authors also showed an increase of the negative
balance of the current account of balance payments increased the probability of devaluation.
       For the analysis Frankel and Rose (Frankel and Rose, 1996) used data from
approximately 100 countries from 1971 t o 1992. They define devaluation as nominal
depreciation of the exchange rate on no less than 25% with the increase of depreciation speed in
10%. Further they develop multidimensional probit-model. Most of the variables responsible for
the structure of the foreign debt turned out to be insignificant.
       Research of Sachs, Tornell, and Velasco (Sachs, Tornell, Velasco, 1996) was dedicated
to the Mexican crisis in 1995. For identification of the cases of “credit booms” authors used
relations of bank claims to the private sector to GDP (B/GDP). If fractional variations of this
variable (LB) were high, then it was considered that “credit boom” took place in the country.
Besides, authors calculated for all country relations of the monetary index M2 to gold and foreign
currency reserves (M2/R) which was interpreted as an indicator of sufficiency of reserves, and
index of revaluation of the real exchange rate (RER) equal to the change of average value of the
real effective exchange rate index in 1990-1994 in comparison to 1986-1989.
       All observations are then divided in periods during which countries had strong
fundamental variables, and periods during which countries had weak fundamental variables.
Authors thought that countries which had strong fundamental variables are the index value of
revaluation of the real exchange rate (RER) at the present moment situated in the upper quartile,


                                                                                                    35
and value of the “credit boom” index (LB) is in the lower quartile. For taking into account the
“quality” of fundamental variables of the country during evaluation of the regression model,
authors introduce fictitious variable equal 1 if the country during the present period has
weak fundamental variables and 0 in the opposite situation. Besides, during the evaluation
process they use variable which characterizes gold and foreign currency reserves and equal to 1
for those countries whose indicator’s M2/R value at the present moment is situated in the lower
quartile, and 0 in the opposite situation. After that econometric model was evaluated (12):


IND  1   2 RER   3 LB   4 D LR RER   5 D LR LB   6 D FW D LR RER   7 D FW D LR LB   .

(12)
       Evaluation results of the equation (14) show that speculative pressure on the exchange
rate of the national currency increases with decline of the gold and foreign currency reserves and
increase of the bank claims to the private sector. Proof of the hypothesis that excessive capital
inflow increases crisis possibility was not found.
       In their study Corsetti, Pesenti and Roubini (Corsetti, Pesenti, Roubini, 1998) examined
Asian crisis. Excessive foreign debt of the Asian companies and deficit of the balance of
payments are named as the main causes of the crisis. Authors pay special attention to the
problem of the moral hazard: Following signaling indicators were used in the article:
   1) Index which reflects crisis probability presents itself weighted average value of monthly
       increases of the exchange rate and gold and foreign currency reserves.
   2) “Health” variables of the financial system is the relation of non-operating loans to the
       total assets of the banks. Other indicators equal to the growth ratio of the relation of bank
       credits to GDP. The third variable takes in account loans issued by the commercial banks
       to non-banking sector.
   3) Index of overs-and-shorts of the current account positions. If growth ratios of the real
       exchange rate on the end of 1996 in comparison with the average valued of 1988–1990
       exceeds 10%, then index equals to current account (% to GDP); in the opposite case it
       equals 0.
   4) Fundamental variables are the relation of indexes M1, M2 and maintenance expenditures
       on foreign debt to gold and foreign currency reserves.
       Further, authors designed a probit-model that reflects dependence of the crisis probability

                                                                                                        36
from the noted variables. Evaluation of the model shows that crisis probability is strongly
influenced by the fundamental variables and imbalance of the balance of payments.
        Demirguc-Kunt and Detragiache (Demirguc-Kunt, Detragiache, 1998b) study banking
crises both in developed and developing countries during 1980–1994. For the econometric
evaluation of the banking crisis probability authors use the multidimensional logit-model.
        In the article, banking crisis is defined as a situation when one of the following cases
takes place (depending on the available statistical data): share of non-operating assets in total
assets is over 10%; costs on reconstruction of the banking system is at least 2% of GDP; and the
problems of the banking sector lead to full scale nationalization of banks.
        Evaluations of the model show that slowdown of the GDP growth rates and a worsening
of the terms of trade increase the possibility of the banking crisis. Rise of the interest rate, relation
of money supply M2 to the gold and foreign currency reserves and speeding up of inflation also
have statistically significant positive correlation with the crisis probability.
        Empirical evidence was found of the hypothesis that countries where banking sector
lends more credits to the private sector will face a higher probability of a banking crises.
        Hardy and Pazarbaioglu (Hardy, Pazarbaioglu, 1998) also devoted their research to the
determination of the role of macroeconomic factors in the process of forecasting financial crisis
probability. They study the role of macroeconomic factors in creation of prerequisites for the
occurrence of the crisis in the banking system. For their analysis authors use data on banking
crises in Asian countries and Latin America of 1 9 80 – 90’s on the basis of which they build a
logit-model. Authors determine three groups of explanatory variables:
       Variables of the real sector:        growth ratio of real GDP; growth ratio of private
        consumption; and the growth rate of private fixed investments.
       Variables of the banking sector: debt on private deposits in relation to GDP; relation of
        issued bank loans to GDP; and the relation of banking sector foreign debt to GDP.
       Shocks that influence banking sector: inflation; deposit interest rate in real terms; real
        exchange rate; growth ration of import in real terms; and the terms of trade.
        Possibility of problems in banking sector strongly depends on GDP growth ratio;
consumer boom might cause crisis; speed up of inflation with its following drop is one of the
most important variables in forecasting banking crisis; bank deposits in real terms decrease both
during the period before the crisis and during the crisis itself; amount of loans to the private

                                                                                                      37
sector sharply increase before the crisis and decrease during the crisis itself; increase of the
exchange rate and interest rate increase possibility of crisis; and the inclusion of the variables,
which take in account regional differences, in the model increases its prognostic value.
        Kruger, Osakwe and Page (Kruger, Osakwe, Page, 1998) analyze factors which
influence occurrence of currency crises in Asia, Latin American and Africa during 1977–
1993. Probit-regression is used for econometric evaluation. For identification of crisis authors used
index equal weighted average value of increases of nominal exchange rate and gold and foreign
currency reserves. If value exceeded a standard deviation of 1.5, then crisis takes place. In the
model, the variable which is responsible for the contagion effect is ( R (Crisis j ,t )) . It takes value

1 if some other country experienced crisis during the same period, and that country is from the
same geographical region as the examined one.
        Authors used following exogenous variables: relation of foreign debt to GDP; relation of
money supply M2 to gold and foreign currency reserves; relation of current account position of
the balance of payments to GDP; real exchange rate; relation of budget deficit to GDP; domestic
credit; GDP growth ratio per capita; inflation; and world interest rate.
        The article shows that the slowdown of the GDP growth ratio, unsecured by gold and
foreign currency reserves, the increase of money supply and appreciation of the national
currency facilitate crisis events. Variables related to foreign debt turned out to be insignificant.
And finally, the contagion effect was evidenced empirically.
        The subject of study of Milesi-Ferretti and Razin (Milesi-Ferretti, Razin, 1998) is a sharp
increase of the current account of the balance of payments and depreciation of the national
currency in the countries with average and low income during 1973–1994. Authors considered
as crisis events following situations during which following conditions were fulfilled
simultaneously:
    a) decrease of relation of current account deficit to GDP in approx. 3% (average value three
        years before the crisis in relation to average value three years after the crisis);
    b) minimal value of current account deficit after devaluation is no higher of its maximum
        value during the period of three years before devaluation;
    c) after devaluation current account deficit decreases on approx. 1/3.
        Analysis was conducted with help of following variables:
       Variables of real sector: GDP growth ratio; growth ratio of consumption; growth ratio of

                                                                                                      38
       investments; budget deficit/surplus; and GDP per capita.
      External variables: current account of the balance of payments; real effective exchange
       rate of nation currency; relation of gold and foreign currency reserves to import; level of
       trade freedom (relation of the sum of export and import to GDP); and international
       transfers (% GDP).
      Foreign debt: relation of foreign debt to GDP; maintenance of foreign debt in relation to
       GDP; share of soft foreign debt in total foreign debt; share of short-term foreign debt in
       total foreign debt; share of state foreign debt in total foreign debt; and the relation of
       attracted foreign direct investments to foreign debt.
      Financial variables: relation of money supply M2 to GDP; growth ratio of domestic
       credit;and the relation of bank loans issued to private sector to GDP.
      Exogenous variables: USA interest rate; GDP growth ration in countries of OECD; and
       the terms of trade.
      Fictitious variables: regional variables; variables that take in account exchange rate
       regime; and the variable that takes in account participation in the IMF programs.
       For evaluation purposes probit-analysis was used, in other words crisis probability was
evaluated in dependence from the values of mentioned earlier indicators with the time lag in one
month. Model evaluation showed positive correlation of the crisis probability with high GDP per
capita, high current account deficit of the balance of payments, low growth ratio of investments,
poor terms of trade, high US interest rate, and stable and fast economic rates of growth in the
OECD countries. Moreover, significant relation of gold and foreign currency reserves to import
and large share of soft foreign debt in total foreign debt negatively correlated with the crisis
probability.
       In their work, the authors also examined currency crises for identification and use the
following four definitions:
   a) currency depreciation in relation to the US dollar on no less than 25% a year, with that
       noted value has to be at least 10% higher of its value of previous year;
   b) currency depreciation in relation to the US dollar on no less than 25% a year, with that
       noted valued has to be 2 times higher its value of previous year; currency depreciation
       for the previous year shouldn’t exceed 40%;
   c) currency depreciation in relation to the US dollar on no less than 15%, with that noted

                                                                                               39
       value has to be at least 10% higher of its value of previous year; currency depreciation
       for the previous year shouldn’t exceed 10%;
   d) in addition to the previous condition, it is needed that a year before the crisis the
       exchange rate was fixed.
       To the list of variables used by authors for the analysis of crises of balance of payments
was added relating currency reserves to import or money index M2 so they could examine
currency crises as well.
       Evaluation of the probit-model showed that increases of the current account of the
balance of payments, decline of gold and foreign currency reserves, over-evaluated exchange
rate and poor trade conditions increase devaluation probability. Slow GDP growth abroad and
high world interest rates also increase probability of the currency crisis, though high international
transfers decrease its probability.
       The aim of research conducted by Deutsche Bundesbank (Deutsche Bundesbank, 1999)
was to gain an understanding of the Asian crisis of 1997. Determination of the crisis events in
the research was made with help of index that took in account changes of the exchange rate and
gold and foreign currency reserves. Interest rate was not taken in account because authors found
its high correlation with the exchange rate. They defined as a crisis a situation when index value
exceeded 1.5 standard deviations.
       The binominal model was established using variables to determine the probability of the
financial crisis in relation to real exchange rate, export, relation of current account of the balance
of payments to GDP, gold and foreign currency reserves, domestic credit, difference in inflation
in the country examined and the USA, and world interest rates. All coefficients of the model
turned out to be significant and had an expected sign. Results confirm that higher interest rate
and slow economic growth in industrial countries increase crisis probability.
       Glick and Moreno (Glick, Moreno, 1999) examined crises in Asia and Latin America
during 1972–1997. They defined a crisis a situation when deviation of the exchange rate from the
average level exceeds two standard deviations. For the statistical analysis probit-regression is
used. They showed that crisis probability increases with the decline of gold and foreign currency
reserves and increase of relation of money supply M2 to gold and foreign currency reserves.
Decrease of the domestic credit also increases devaluation probability.
       Tornell (Tornell, 1999) examined Asian and Mexican crises in a similar way. He


                                                                                                   40
studied countries with the developing economies only, and suggested that investors will exercise
speculative attacks on the currencies of those countries where devaluation probability is high, i.e.
countries with highly over-evaluated currencies and small gold and foreign currency reserves.
       “Credit boom” in Tornell’s work is identified with help of the variable LB equal to
increase the volume of loans issued by banks to the private sector and government corporations
during previous four years in real terms. Variable RER presents itself a change of real effective
exchange rate during previous four years. Besides, in econometric analysis fictitious variable
DHR is used. It equals 1 if relation of money supply M2 to gold and foreign currency reserves
prior to crisis was less than 1.8, and 0 in the opposite situation. Finally, author divides
observation on the periods when country had strong fundamental variable (LB<0%, RER<5%)
and all other cases. In the first case value of the fictitious variable was equal 1, and 0 in the
opposite situation. After that equation (13) was evaluated:


IND  1   2 RER   3 LB   4 D LR RER   5 D LR LB   6 D FW D LR RER   7 D FW D LR LB   .

(13)
       Evaluation results show that in countries with low gold and foreign currency reserves and
weak fundamental variables, devaluation and “credit boom” increase significantly crisis
probability, and on countries with strong fundamental variables exchange rate and “credit boom”
do not influence crisis probability. The article also shows that countries from the same region as
the source of crisis, suffer from it more.
       In the research of Carramazza, Ricci and Salgado (Carramazza, Ricci, Salgado, 2000)
the applicability of financial variables for forecasting of crisis events is examined, attention is
also paid to the role of the institutional factors. Countries influenced by Mexican and Russian
crises are examined only.
       Speculative pressure index is calculated during the first part of crises identification. On the
second step authors evaluate probit-model of regression on panel data from 1990 t o 1998.
Further, authors introduce the concept of “common creditor”, i.e. country which lend the largest
amount – source of crisis. Variable BISB shows importance of the common creditor for the
country, it equals to the share of credits issued to the country by common creditor from total
amount of credits issued through the Bank for International Settlements (BIS). Variable BISA
shows the importance of common creditor for the country (it is a share of borrowings received

                                                                                                        41
from the common creditor in all borrowings of the country). Then, uniting these variables it is
possible to establish new index BISAB equal to the product of BISA and BISB .
       As exogenous variables, the authors used two groups of variables. The first group
included variables whose influence on the crisis occurrence can be explained theoretically: real
exchange rate, current account of the balance of payments in relation to GDP, export share in
GDP, budget deficit (% GDP), relation of M2 to GDP; GDP growth ration in real terms,
contagion index.
       Contagion index includes two components. The first component allows taking into
account income effect and equals to weighted average change of industrial output of the main
trade partners of the country during the year after the crisis. Price effect recognizes change in
competitiveness and is calculated as growth rate of real exchange rate of the country during six
months after the crisis.
       The second group of variables describes capital outflow and weakness of the market:
share of short-term foreign debt in relation to the total debt, variable BISAB ,volatility of the
stock market index, correlation of stock market index of the country and country the source of
crisis ,relation of M2 to gold and foreign currency reserves, relation of short-term foreign debt of
the country to gold and foreign currency reserves.
       Model evaluation allowed researchers to find significant positive correlation with the
probability of the crisis of the real exchange rate growth, a growing current account deficit of the
balance of payments, a slowdown of output in real terms, the growth of the share of short-term
foreign debt in relation to the total foreign debt, and also the presence of the common creditor
(BISAB), increase of the relation of short-term foreign debt of the country to gold and foreign
currency reserves and relation of M2 to GDP.




2.3. Non-parametric evaluation


       The first work in which a non-parametric evaluation approach was used to forecast
financial instability was the research of Kaminsky, Lizondo and Reinhart (Kaminsky, Lizondo,


                                                                                                 42
Reinhart, 1998). They performed empirical analysis of the currency crises in 1990’s and offered
their own system of early alarm signals to determine approaching critical events.
         The authors defined as a crisis a situation in which a speculative attack on a currency
leads to its fast devaluation, decrease of gold and foreign currency reserves or combination of
two of these factors. Identification of these crises was made with help of the speculative pressure
index equal to a weighted average change in currency exchange rate and gold and foreign
currency reserves in a month.
         In their research authors study following indicators: gold and foreign currency reserves; imports;
exports; terms of trade; real exchange rate; and the spread between world and domestic interest rates;
excessive money supply in real terms; monetary multiplier; relation of domestic credit to GDP; real
interest rate on deposits; relation of the credit interest rate to deposit interest rate; deposits of the
commercial banks; GDP in real terms; and the stock market index.
         As signal horizon, or the period within which dynamics of indicators may forecast crisis,
24 months was used. For each of the indicators in each country, a threshold is defined separately.
If the indicator’s value exceeds the threshold, then it means it signals. Threshold levels were
selected so that on the one hand indicators would not give many false signals, and on the other
hand would not miss the critical events.
         Each indicator may signal (first row Table 1) or may not signal (second row). If
indicators signal, if afterwards a crisis follows within the 24 month time horizon, then the
signal is “good” (cell A). If indicators signal and a crisis does not happen within 24 months, then
the signal is considered as noise or “bad“ signals (cell B). If the indicator does not signal and a
crisis happens, then the signal considered “missed” (cell C). If the indicator does not signal and a
crisis does not happen within 24 months, then signal is also considered as a “good” one (cell D).
Ideal indicators will be characterized with non-zero values only in cells A and D.
                                                                                                  Table 1
                            Distribution of indicator’s signals (Kaminsky, Lizondo, Reinhart, 1998)


         Event                 Crisis within 24 months                  No crisis within 24 months
Signal                                     A                                          B
No signal                                  C                                          D




                                                                                                        43
        The number of months is placed in the cells of Table N during which each event took
place. During the selection process of indicators, the share of “good” signals A/(A+C), “bad”
signals B/(B+D), and “relation of noise to signals” [B/(B+D)]/[A/(A+C)] were taken in account.
Besides, for the “good” indicator conditional probability that crisis will happen A/(A+B) has to
be higher than unconditional (A+C)/(A+B+C+D).
        In their work, the authors showed that the signaling approach might be effective at
forecasting critical events and indicators in which prognostic values were found: exchange rate,
domestic credit, money supply, gold and foreign currency reserves and export.
        Later Kaminsky developed a non-parametric approach to forecasting financial crises. In
her work (Kaminsky, 1999) she studied currency and banking crises of the 1990’s. In that work
the time interval is divided between “calm” and crisis periods. First, it is being tested if the
numbers of signals during the crisis are higher than number of signals during the “calm” period,
then indicators behavior is examined in relation to the approaching moment of crisis. After that,
signals are divided on “soft” and “rough” (depending on how much they exceeded threshold
levels) and examined separately.
        It is noted in the research that if indicators suggest a crisis immediately before one
occurs, then most likely it shows the happening of the event instead of forecasting it. That is why
the author studied the time structure of the signals. It turns out that increases of the total number
of signals before the crisis is not that significant and average number of signals during the past
six months and previous months do not differ much. Thus, “good” indicators are equally good at
their job as right before the crisis and some time prior to it.
        Kamisnky’s research was one of the first where the attempts of composition of
consolidated indexes. In the article for n of available indicators of financial stability a few
variants of such indexes were reviewed:
                                   n
       sum of all signals: I t1   Sti , where Sti  1 , if at the moment of time t indicator gave a
                                  i 1

        signal i;
               n
       I t2   ( SM ti  2SEti ) , where SM ti  1 , if at the moment t indicator i gave “soft” signal
              i 1




                                                                                                         44
           and SEti  1 , if at the moment t indicator i gave “rough” signal4;
                  n
          I t3   Sti s , where S ti s  1 , if indicator i gave signal during the time line from (t-s) to t;
                 i 1

                  n
          I t4   Sti / wi , where w i is a relation of “noise” to signals of indicator i.
                 i 1

           Further for each indicator and each t=1…T is possible to determine conditional
probability Pt of the event Ct, t+h, crisis will happen during the time period of [t; t+h] at the
condition that there was a signal:
Pt k  P(Ct ,t h I  I t  I 

= periods when I  I t  I and crisis happens within h month / periods when I  I t  I

           Further Rt =1 is being set for t=1…T if crisis actually happens during the time period of [t;
t+h] and Rt =0 in the opposite case. Assessment of accuracy of forecasting of all four indexes
showed that the best forecast is made by index I4. Besides author discovered that the cumulative
index allows for better forecasting crises than separate indicators.
           In their work Kaminsky and Reinhart (Kaminsky, Reinhart, 1999) continue to examine
banking and currency crises in industrial and developing countries, which happened during the
period of 1970-1995. Specifically they pay attention to the situations when currency and banking
crises happen simultaneously, i.e. so called “crises-twins” take place. Identification of the
currency crises is made with help of the index equal to weighted average value of the
fluctuations of the exchange rate and gold and foreign currency reserves. By the beginning of the
banking crisis, it was considered a case of massive withdrawal of deposits, or merger or
acquisition of one or more financial institutes by the government.
           For the analysis of the currency rises the signal window was chosen starting with 24
months before the crisis and finishing with its start; for banking crises signal window included
12 months prior to crisis and 12 months after.
           In that article authors test the same indicators as were used in the previous studies.
Threshold levels are also selected in such a way that relation of the noises to the “good” signals
was minimal. Conducted analysis shows that variables characterizing capital movements are
good to predict currency crises. Variables of the real sector are more useful for forecasting

4
    “Rough” threshold values increased in comparison with the “soft” values on some exogenous set value.

                                                                                                              45
banking crises.
       Edison (Edison, 2000) also makes an attempt to establish a system of early alarming
signals which could be useful in forecasting financial crisis. His judgments mostly comes
works by authors Kaminsky, Lizondo and Reinhart (Kaminsky, Lizondo, Reinhart, 1998),
which were broadened in few directions. Edison examines following indicators:
      Indicators of current account positions: real exchange rate, import, export.
      Indicators of capital transactions account: gold and foreign currency reserves, relation
       of the money supply to gold and foreign currency reserves; spread between domestic and
       world interest rate.
      Indicators of real sector: GDP in real terms, stock market index.
      Financial indicators: monetary multiplier, relation of domestic credit to GDP, real
       interest rate on deposits, relation of credit interest rate to deposit interest rate, excessive
       money supply, deposits of the commercial banks.
        Edison analyses financial crises from 1970−1998. He established the same threshold
levels as Kaminsky and her colleagues. With that, he showed analysis of the dynamics of the real
exchange rate, export and relation of money supply to GDP are better to forecast financial crisis.
       An alternative approach on the definition of threshold levels (1.5 standard deviations
from the average value) was offered which allows the determination of a few well working
indicators. But as a whole, results were significantly worse than at defining threshold levels on
the base of the “noise” minimization to the “good” signals.
       Besides, Edison used consolidated indexes used by Kaminsky (Kaminsky, 1999). His
analysis showed that in many countries these indexes actually increase before the crisis. He
concluded that it is impractical to fully rely on such approach because of the large dispersion
of indexes and problems of their interpretation.
       Hawkins and Klau (Hawkins and Klau, 2000) also make an attempt to establish
consolidated advancing indexes of financial stability. For forecast of approaching financial
instability authors suggest to use three consolidated indexes: speculative pressure index, index of
eternal exposure, and index of the exposure of the banking system. Methodology of the index
calculation is presented in detail in Appendix 1.
       The authors analyzed indexes dynamics of those countries which suffered from the
Russian crisis, crises in Asia and Latin America and showed that examined indexes allow

                                                                                                         46
making statistically valid forecast of financial instability.
        Thus, in this chapter, I reviewed the main approaches of examination of indicators
which allow a signaling before approaching financial crises.
CHAPTER 3


Development of the framework of signaling indicators


        Literature review from the first two chapters helped to determine main analysis
approaches of the financial stability indicators and define their threshold levels which would
mean financial instability in the short-run. These approaches include: qualitative analysis,
econometric analysis and non-parametrical methods of analysis which include establishment of
consolidated indicators
        Qualitative analysis is conjugated with substantial subjectivity in interpretation of the
dynamics of the signaling indicators. That is why to my opinion it is necessary to develop some
quantitative characteristics which would facilitate the monitoring of the financial stability to
more objective one. Literature review on the subject allows the argument that there are two main
ways of creating of such characteristics – econometric modeling and non-parametric evaluation.
        Econometric modeling is based on the scoring models of binary selection with different
indicators of financial instability used as exogenous variables. In this paper I discard econometric
analysis due to the following reasons.
        First, in comparison with non-parametric evaluation methodology of econometric analysis
is significantly more complicated and requires implementation of the large number of theoretical
prerequisites in respect of the origins of the data used. At the same time methodology developed
in this paper is relatively simple and its results are easy to interpret.
        Second, literature review of the applicability of econometric models in terms of
probability evaluation of the financial instability show that in spite of the examination of the same
episodes of the crisis events, results of different authors differ significantly in relation to the
choice of the best signaling indicators and selection of the threshold levels. Choice of a particular
econometric model and interpretation of its results for monitoring of financial stability purposes
will be no more objective than simple qualitative analysis of indicators. None of the scientific
works reviewed demonstrate an advantage of econometric evaluation in comparison to non-


                                                                                                  47
parametric methods.
        Third, by virtue of non-linearity of the models of binary selection it seems to be more
complicated to evaluate the input of each of the regressors in the increase of probability of the
financial instability in the case of actual value of indicator is significantly deviated from the
average value.
        Finally, for obtaining statistically significant results it is necessary to examine bigger
number of the crisis events. In case of Russian Federation, there were only four episodes on
which the statistical data is available. These are: crisis of interbank credit market in August,
1995, the crisis of the stock market in October, 1997, full scale financial crisis in August, 1998
and creditability crisis to the banking system in May, 2004. It is obvious that four crisis episodes
are not enough for the evaluation of the model of binary selection. Model evaluation on panel
data, i.e. with the use of data from the other crisis events in other countries, on my opinion is not
acceptable because it will significantly decrease the strength of a test whereby the crisis
probability is evaluated. In spite of the common features, crises in different countries have many
specific features due to the differences in national economies. That is why the dynamics of the
indicators of financial stability before the crises differ in each different country.
        Therefore, econometric methods of the model’s evaluation are related to the specifications
of the task. They are determined by the following factors:
       Crises in different countries are not homogeneous enough which prevents consolidation
        of previous experiences;
       Each case has its own unique characteristics; some features that can point on the
        vulnerability of economy, cannot be measured mathematically;
       Necessary data is not available;
       With time crises determinants can significantly modify and change.
        Thus, I will adhere to non-parametric methods. Obviously, these methods have their own
limitations. In particular, it is harder to use standard statistical test through implementation of
these methods. Most of the scientific research devoted to forecasting financial instability uses
the so called “signaling approach”, which was first introduced in the research of Kaminsky,
Lizondo and Reinhart (Kaminsky, Lizondo and Reinhart, 1998). That is why in my work I will
make an attempt to adapt “signaling” approach for establishing the framework of indicators of
financial instability on the Russian Federation market.

                                                                                                  48
       Besides the search of indicators dynamics of which is the best way to reflect approaching
financial instability, it is important to develop some consolidated indicator of financial stability,
which would accumulate all information gathered from the separate signaling indicators. In my
empirical part of my work I will analyze the dynamics of different consolidated indexes of
financial stability and will choose those which are the most practical for the analysis of the
Russian financial market.



   3.1. Signaling approach


       The analysis of theoretical and empirical aspects of the interrelation of different macro-
and microeconomic variables and the probability of the financial system crisis allows for the
identification of a series of variables which can be used as the signaling indicators:
      Rate of economic growth: GDP growth ratio, dynamics of industrial output
      Balance of payments: current account balance, gold and foreign currency reserves,
       foreign debt, terms of trade (export prices), export and import, real effective exchange
       rate, capital outflow.
      Interest rates: real exchange rate, difference between world and domestic interest rates
       in real terms, relation of credit rate to deposit rate.
      Monetary indicators: customer price index, dynamics of the domestic credit in real terms,
       monetary multiplier, relation of monetary supply to gold and foreign currency reserves,
       growth ratio of deposits in real terms, excessive money supply in real terms, speculative
       pressure index
       It should be noted that in practice not all indicators can be used for the analysis of the
financial system stability because statistical data on some variables may not be available. Each
approach to the analysis of indicators should be subjective, i.e. with the use of expert estimation
and information about market conditions which cannot be presented in terms of quantity
variable. That is why a distribution of indicators for different sectors of the financial market is
relatively conditional and most of attention should be paid to the overall economic condition in
the country.
       Literature overview shows that financial crises are usually preceded by negative trends of


                                                                                                  49
the main macroeconomic indicators. Therefore, a system of signaling indicators should include
variables that allow estimating the overall situation in the country. When available statistical data
on the separate sectors of the financial market is limited by a small number of indicators, the
variables obtained have especially significant meaning. Let’s take a closer look at these
indicators.
       Economic growth. Rate of economic growth is the key index of economic dynamics,
which permits judgment of how successfully the economy will expand. GDP growth ratio in
real terms and dynamics of industrial production are suggested to use as main signaling
indicators.
       It is assumed that a slowdown of economic growth reduces the ability of national
borrowers to pay off their debts and therefore increases credit exposure. Recessions often precede
major financial crises.
       Balance of payments indexes. Main indexes of balance of payments provide details
which enables a timelier manner of information about an increasing probability of currency crisis.
This also allows the following up of information leading to approaching external shocks.
       In particular, an increase of ratio of current account balance to GDP usually leads to
significant export revenues inflow in the country which gets absorbed by the financial system. At
the same time, substantial deficit on current account may provide a signal about increasing
probability of a currency crisis and reduction on liquidity of the financial system. In turn, an
increase of exchange risks is able to cause short-term investment outflow and worsen financial
instability. Moreover, a downswing of gold and foreign currency reserves or increase of
foreign debt is also considered as an obvious sign of financial instability.
       Scientific evidence justifies that substantial deterioration in terms of trade leads to
difficulties in the financial sectors in many countries. Small economies with heavy dependence
from the export of raw materials and resources are the most vulnerable to shifts in world market
conditions. Real effective exchange rate is closely related to the terms of trade; its increase leads
to deterioration of competitiveness of domestic producers and might lead to a slowing down of
the rate of economic growth or recession.
       Moreover, substantial acceleration of capital outflow may signal approaching
instability, which causes pressure amplification on the exchange rate of the national currency.
       Finally, export and import dynamics are used as stability indexes of the balance of


                                                                                                  50
payments. Usually currency crises were preceded by a reduction of exports and increase of
imports.
       Interest rates are the fundamental characteristics of the financial market. Understanding
their dynamics provides monitoring towards the stability of the financial system and notifying at
a reasonable time about emerging problems.
       Real interest rate is the most important index of the group. An increase in the real interest
rate results in an increase of instability of the financial system and encourages growth of
coefficient of non-working debt. At the same time, a sustainably negative real interest rate speaks
about existing disproportions in the financial system. One reason might be an attempt of the
government to fixate nominal interest rate. It makes sense to analyze the volatility of the interest
rate along with its level. An increase of volatility of the interest rate speaks about increase of
interest rate risk; and therefore, about increase of instability of the financial system.
       Besides national interest rate, international interest rates also play an important role for
the financial system. Increase of the world interest rate increases vulnerability of the national
financial system. This is because capital outflow from the developing markets to developed
countries takes place, worsening the creditworthiness of borrowers on the developing markets (at
foreign currency loans). That is why in the number of signaling indicators of financial crisis
included spread between domestic and foreign interest rate.
       Finally, the relation of credit rate to deposit rate is being analyzed. Before currency crises
start enlargement of spread between credit rates and deposits was observed in many of the cases.
It is because domestic credit growth precede currency crisis. In this situation the share of “bad”
credits grows and banks raise credit rates trying to compensate possible losses from outstanding
loans. Deposit rates also rise but in a lower proportion.
       Monetary indicators. Analysis of monetary indicator dynamics might be extremely useful
in forecasting crisis in the financial system. Thus, acceleration of rate of growth of consumer
prices before a crisis is possible. Besides, rapid growth hinder estimation of the credit risk and
enhances uncertainty. Sharp drops of inflation can also lead to the decrease of nominal yield and
cash flows, which can erode stability of the financial institutes.
       Financial crisis are often preceded by the expansion of the domestic credit, including an
increase of the share of “bad” loans. That is why relation of domestic credit to GDP is included
in the list of possible signaling indicators. Excessive money supply in real terms (considerable


                                                                                                  51
extent of which speaks about increased possibility of the financial crisis) is defined as a
deviation of weighted money demand from observed money supply (expressed as a share of the
money quantity in GDP).
           One of the most important indicators in this group is the relation of deposits to the money
supply. Decline of this relation index might signal a loss of creditability to the banking system
and as such lead to occurrence of liquidity crisis. At the same time, a decline of the indicator may
signal that financial institutions other than banks are more effective.
           The relation of credits to deposits can illustrate the ability of the banking system to obtain
the means necessary for the demand satisfaction on loans. High value of the index might justify
problems in the banking system and low level of liquidity.
           Finally, the monetary multiplier also is able to signal instability of the financial system.
Considerable growth of a multiplier might be the sign of weakening of selection procedure of
fund receivers by the commercial banks. At the same time, in many of developing economies,
including Russia, this index is still on a relatively low level in comparison with the developed
countries, that is why its dynamics should be interpreted carefully.
           Speculative pressure index was included in the list of signaling indicators for the
monitoring purposes of the situation on the exchange market. This index is an aggregated index
and allows evaluating stability of the exchange rate of the national currency in a short-run5.
           Such index presents a weighted-average of the three variables:
                                                                                 
       1) Increase rate of the exchange rate of the national currency per month, E ;
                                                                                  
       2) Increase rate of gold and foreign currency reserves (reversed in sign), R ;
       3) Level of the interest rate (for Russia it is a weighted-average interest rate on the ruble
           credits to corporate entities in the lending institutions), i.
           Two last variables reflect fiscal and monetary policy of the governmental bodies on the
exchange market in case of the speculative attack on the exchange rate of the national currency.
           It is assumed that at the fixed exchange rate speculative attack on the exchange rate will
appear as a decrease of the gold and foreign currency reserves. At the same time at any given
regime of the exchange rate, for the protection of national currency the central bank may raise
interest rates. It is taken in account through including in the formula to calculate third variable.
           Thus, speculative pressure index is calculated in a following way:

5
    This index was first introduced in the article of (Eichengreen, Rose, Wyplosz, 1995)

                                                                                                       52
                         
             w1 E  w2 ( R)  w3i
        I                                                                            (14)
                       3
        Weights w1 , w2 , w3 , are picked up so that dispersion of all three values were equals, in

                                                                                   
                                                                                D( E )             
                                                                                                D( E )
                              
other words, D( w1 E )  D( w2 R)  D( w3 i ). Therefore, taking w1  1, w2           , w3           .
                                                                                   
                                                                                D( R)           D(i )

Statistical methods of time series analysis are used for the composition of this indicator.
        Researchers who offer formulas for the calculation of this index considered two threshold
levels, exceeding any of each was assessed as a thread of the currency crisis: mean value of
index through the entire period and value equal to three standard deviations.
        In most of the cases I use growth ratios of the variables or examined their relation to GDP
to ensure the comparability of data. In some cases variables described in levels were used
because in such form they show the most prognostic power. Table 4 (Appendix 2) presents the
description of the transformation of the signaling indicators, report frequency and source of
information.
        Further of the analysis of the signaling approach I will observe some methodological
matters related to it.
        Monitoring instruments. Main advantage of the “signaling approach” is that assessment
of the predictive value of each of indicators is proceeded individually, which allows to range the
variables. Moreover, this methodology can be used for the development of the present economic
policy, because the variable that signals might be defined precisely. Risk probability in this
methodology is represented as a binary function of the indicators value, which takes 0 value
when indicator variable is less than threshold level and it takes value 1 in opposite case. Thus,
this model does not recognize cases when variable exceeded threshold insignificantly and when
it is significantly more than a threshold level.
        Thus, the offered system of indicators is mostly based on the “signaling approach”. At
the same time, statistical and econometric methods are used to compose certain indicators. In
particular, according to Kaminsky, Lizondo, Reinhart (Kaminsky, Lizondo, Reinhart, 1998)
methodology, excessive real money supply (significant value of which signals about high
possibility of the financial crisis) is defined as a deviation of weighted money demand from
observed money supply (expressed as a share of the money quantity in GDP), i.e. as the remains
of the regression equation of the following form:

                                                                                                     53
         Mt
              a 0  a1Yt  a 2 pt  a3 t   t ,
        GDPt
                                                                                     (15)
where, Mt - money supply M2; GDPt – nominal GDP; Yt – GDP in real terms;  p t - consumer

                                             Mt
price index; t – time. In this equation          reflects observed money supply, while expression
                                            GDPt

from the rights reflects money demand. Thus, remains  t are interpreted as the variable of

excessive money supply, i.e. difference between money supply and demand.
        Monitoring frequency. Signaling approach describes signal as an exit of one of
indicators threshold levels. If indicator signals during the certain period of time before crisis, so
called “signal window”, then such signal is considered “good”. If indicator signals and critical
events do not happen during the certain period of time, then this signal is considered “bad”.
        The most preferable monitoring frequency is set up exogenously and signal window in 3
months before the crisis. A large signal window is not practical because situations in the
financial market are highly volatile and change quickly. In other words, I believe that negative
tendencies which can potentially lead to the financial instability are possible to determine 1-2
quarters before their occurrence. Besides, if these negative tendencies arise earlier, then closer to
the potential date of financial instability they will only get stronger and at the same time noted
period of time is relatively enough to neutralize negative influences. Thus, quarter monitoring
allows timelier notification of negative tendencies in the financial system of the country and
permits the taking of preventive measures to ensure financial stability.
        Moreover, quarters are the optimal window because data for most of variables is not
published more frequently than this. Thus, I will analyze indicators of the financial stability on a
quarterly basis, though for the maximum revelation of their dynamics I will use monthly statistic
data if available.
        What concerns time series necessary for the accumulation of the critical changes of each
of the signaling indicators, I assume that such period is two-three quarters. Such conclusion is
drawn from the scientific evidence. Unfortunately, like in many other cases while creating
system of indicators it is not possible to evaluate this period quantitatively because of the
absence of the necessary statistic data.
        Threshold values. Signaling approach assumes the necessity of the test of the zero
hypothesis that economy is at its normal stage of development against alternative hypothesis that
                                                                                                  54
within 3-6 months it will experience financial instability. Just like with testing any statistical
hypothesis it is necessary to select some critical value which divides indicators distribution on
two zones6. If indicator’s value hits critical zone, exceeds threshold level, then it is considered
and noted indicator signals.
         The question of defining the level of indicators typical for the normal functioning of the
financial system is very difficult. It also concerns threshold values exceed of which will allow to
talk about high probability of the financial crisis in a short-run. The situation is possible when
the same dynamics of the indicators will signal equally about increasing probability of the
financial crisis and normal development of the financial system at the same time. Analysis each
of indicators and explanation of its dynamics separately seem to be the most sensible approach in
this case. Then obtained information will be put together drawing an overall situation in the
economy and finally the conclusion about high/low probability of the financial crisis is made.
         I will analyze indicator’s dynamics and conclude that variable’s behavior shows high
probability of the crisis in a short-run if there will be an observation of high increase/decrease of
the variable’s value (in dependence from the theoretical hypothetic behavior of the indicator
before the crisis).
         Threshold levels are defined in such way so that on the one hand they maximize the
amount of “good” signals, and on the other hand, minimize “noise” signals, i.e. when an
indicator signals and nothing happens within “signal window”. In that case if it is not possible to
set up a threshold level for an indicator that would ensure acceptable level of “good” signals with
the set level of “noise”, such indicator will be excluded from examination.
         For the optimal choice of threshold level for each of the indicators it is necessary to set
up some criteria. Relation of share of “bad” signals to the “good” signals will be criteria of that
kind. To explain this criterion all values of indicator will be divided in four groups presented in
the table below. Obviously, in case of ideal indicator its values will fit cells A and D only.


                                                                                                                  Table 2
                                      Distribution of the indicator’s values during the signaling process
                              There is a crisis in 3 months                           No crisis in 3 months


6
  I will test single-sided hypothesis, i.e. I assume that either increase or decrease of the variable’s value can justify
increasing probability of the financial instability.

                                                                                                                            55
Signal                                A                                         B
No signal                             C                                         D


         Unconditional probability of the financial instability for each of indicators is set as
relation of observations after which financial instability followed within 3 months to all
observations:
                      AC
         P(C )                ,
                   A B C  D                                                       (16)
         If indicator sends big amount of “good” signals, has good working capability, then it can
be expected that probability of financial instability at the condition of the signal P(C | S )
(conditional probability) will be higher that unconditional P(C). With that
                      A
         P(C S )         ,
                     A B                                                            (17)
         In other words, it makes sense to use indicator for forecasting financial instability if
following correlation is fulfilled:
         P(C S )  P(C)                                                              (18)

         This condition is necessary for the selection of the optimal threshold level. Moreover,
the relation of “bad” signals to “good” ones is minimized in the following way:
                   B /( B  D)
         N /S 
                   A /( A  C )                                                      (19)
         Thus, possible threshold levels for each indicator are described and a threshold was
selected at which variable values (18) was minimal and condition (17) was fulfilled. Notably, in
some cases relation of “bad” signals to “good” ones is equal 0 because share of “bad” signals
equals 0 too, however if the indicator is not sensitive enough, i.e. does not signal before many of
critical events. That is why at the selection of indicators and threshold levels it is necessary to
pay attention on the share of crises (PC) which indicator predicts. In other words, to identify if a
given indicator signals within the given period of time before the crisis at least once.
         In the next two sections I will analyze the performance capacity of the offered system of
signaling indicators on the example of the financial system of Russian Federation, establish
indexes of financial stability, and after will conduct an analysis of the financial stability in
former Soviet republics and transitional economies around the globe.

                                                                                                 56
3.2. Analysis of operational capacity of the potential signaling indicators based on the
example of Russian Federation in 1994-2009


        Methodology of signaling approach described earlier allowed receiving of the results
presented in Table A1 (Appendix 1). It was possible to calculate quantitative characteristics for
all indicators, except foreign state debt and dynamics of industrial output which do not have
enough of the statistical information. Threshold values were defined on the basis of the data
analysis from the period of 1994–2009 (with account of their availability on some variables).
         In Table A1 variables are ordered according to their prognostic power, which is in excess
of conditional probability of financial instability over unconditional. In other words, the better
the indicator, the higher the probability of financial instability under conditions of a given signal,
and therefore, excess of this probability over unconditional probability because unconditional
probability does not depend from the choice of threshold value. Differences in unconditional
probability are determined only by the fact that different indicators have different amount of
data.
         Received results allow a conclusion that use of all indicators except net capital outflow
allows the forecasting of financial instability with probability surpassing unconditional.
However, in my opinion it makes sense to set some limit of such excess so it will not be beyond
the scope of statistical error. Such a limit could be the difference between conditional and
unconditional probabilities of financial instability in size of 5%. In that case good indicators will
be first 13 from Table A1.
         In spite of a relatively high working capability of offered framework of indicators,
examination of received statistical characteristics of signaling indicators still can result in a
question of their effectiveness as most indicators predict no more than a half of crisis
episodes; and increase of conditional probability of financial instability in comparison with
unconditional for some indicators does not exceed 5–10%. Of course it is necessary to
understand limitations of the offered framework. In particular, this methodology gives only
certain information about occurring tendencies on the financial market, but does not show that
financial crisis will happen for sure. Analysis of the signaling indicators allows a monitoring of


                                                                                                   57
long-term tendencies in the economy. Separate negative tendencies can be compensated by
advantageous factors in the short-run. At the same time the market slump accumulated negative
events in the economy can cause financial instability in the end.
          Table 3 and Appendix 2 present the condition of the framework of signaling indicators of
financial instability prior to crises events in Russian Federation. These tables are based only on
working indicators, the first 13 from the Table A1.


                                                                                             Table 3
           Condition of the framework of signaling indicators before the crisis episodes in Russian
                                                                                         Federation
          Crisis              Number of indicators    Number of indicators     Share of indicators
                              with statistical data   which signaled during   which signaled in total
                                available prior to    3 months period prior   amount of indicators,
                                      crisis                to crisis                   %
Crisis of interbank credit             12                      5                        42
market in 1995
Crisis of the stock market             13                      6                        46
in 1997
Currency crisis in 1998                13                      9                        69
Creditability crisis to the            13                      5                        38
banking system in 2004



      From Table 7 it is clear that biggest amount of indicators signaled about an approaching
financial crisis in 1998. It is an expected result because the noted crisis event had the largest
scale of all events examined. Around 40% of indicators signaled prior to the rest of the episodes.
          It is possible to see from the conducted analysis in the previous chapters that one of the
most difficult milestones of the construction of the system of signaling indicators is a selection of
the proper variables for the forecasting of crises in each given case. Table A2 of Appendix
shows a number of publications in which indicator was found statistically significant in its
empirical evaluation
          In my work I will perform similar test of the operational capability of indicators but for
larger number of indicators, specific to the recent creditability crisis to the banking system in

                                                                                                  58
2004, currency crisis in 1998, crisis of the stock market in 1997 and crisis of interbank credit
market in 1995. The test outcomes are presented in the table A3 of Appendix.
      Thus, after conducting analysis of the operational capability of the potential signaling
indicators of the financial crisis in the Russian Federation and their behavior before the crisis
events which were observed in the past, the data leads to the conclusion which suggested
indicators allows a determining in advance of symptoms of financial instability. Unfortunately,
there is insufficient statistical data for testing operational capability of many indicators
(especially as it concerns variables, specific for the separate sectors of the financial market of
the Russian Federation). Scientific expertise shows that they also play an important role in
forecasting financial instability and therefore must be included in the offered system of
indicators. Moreover, to forecast the type of the crisis is very difficult, that is why it is needed to
analyze a whole population of offered indicators.
      As it was mentioned previously, during crisis events indicators do not equally work well,
in other words in separate cases we can observe negative tendencies in the dynamics of
indicators, which might not be followed with the crisis events afterwards, so called “false
alarm”. Moreover, the situation is possible when indicators remain stable in spite of an existing
obvious crisis event, or so called “ignored event”, but it is not a sign of their bad quality.
Because of the diverse behavior of indicators through different types of crises it seem to be not
possible to set up some absolute threshold level after exceeding which crisis occurrence become
inevitable. In proportion to accumulation of necessary statistical data the next step of the
development of the system of signaling indicators must be a development of more formal
procedure of indicators selection and evaluation of their quality.
       In regards to an offered system of indicators, before the liquidity crisis in the banking
system of Russian Federation in 2004, the relation of operational indicators to non-operational
was 7:9, which is a relatively good result because it suggested variables are aimed to signal about
the problems in all sectors of the financial system of the Russian Federation and many of them
simply do not fit to forecast critical events specifically in the banking sector. At the same time,
after crisis in 2004 the offered system did not give false signals about the crisis events.




                                                                                                    59
3.3. Composition of the indexes of financial stability

       In the previous chapters 13 signaling indicators of financial stability were selected which
allowed forecasting of the best critical eposides which took place in the Russian Federation.
However, in the practical application of these indicators, problems are identified in their
aggregating of their signals for composition of the consolidated indexes of financial stability. In
this section a few variants of calculation of these indexes will be offered and I will try to select
the best one on the base of defined criteria.
       Let’s define X vector which consists of the values of 13 selected indicators. Previously it
was mentioned that indicator X j signals during the period t (dummy variable S t j takes value 1) if

it crosses defined (in the previous section) threshold value X j :
       {St j  1}  {St j , X t j  X j }
                                                                                    (20)
       In the expression absolute values X are used because values of some indicators decrease
prior to crisis and some increase. In case of no signal:
       {St j  0}  {St j , X t j  X j }
                                                                                    (21)
       Thus, we can examine a few consolidated indexes of financial stability which are based
on the indexes offered in the articles of Kaminsky (Kaminsky, 1999) and Hawkins and Klau
(Hawkins and Klau, 2000) Appendix . First index is the sum of all signals at the moment t:
                13
        I   Stj
         1
         t
                j 1
                                                                                    (22)
       It is clear that this consolidated index does not take in account many factors. For example,
financial instability probability may increase but it does not mean that all indicators will signal
simultaneously. That is why following a graduate accumulation of the problems in economy the
following index will be used:
                13
        I t2   S t j s ,t
                j 1
                                                                                    (23)
where, St s,t equals 1 if j indicator signals at least once within S months before the moment t.

                                                                                                 60
Parameter S is set exogenously based on our assumption that crisis symptoms are supposed to
reveal a minimum of 3 months before its start; so it is assumed that S equals 3.
        Both of the described indexes are not used to the fullest information received during
setting threshold values of the signaling indicators, because they do not take into account the
forecast precision of each of indicators. The logical approach of accountability of such
information is an indicators’ weighting with help of values equal excess of conditional
probability of the financial instability over the unconditional in case of the signal:
               13
        I t3   S t j ( P j (C S )  P j (C ))
               j 1
                                                                                         (24)
        Notably, Goldstein, Kaminsky and Reinhart (Goldstein, Kaminsky, Reinhart, 2000) use
as weights values opposite to relations of “bad” signals over “good” one for reach of the
indicators. However some of these values equal 0 that is why I use alternative weights. In my
opinion, the better analysis of the dynamics of indicator allows predicting financial instability in
comparison to unconditional probability of the financial crisis (for example, we can say that
evidence shows that crisis happens in average once in 3 years), then higher weight it should
have.
        After calculating the value of each of the three indexes during the certain period of time,
we can calculate their threshold values the same way as it was done for the signaling
indicators. However, such threshold values are able to receive binary information only. In other
words, if the index hits its threshold value it means that with high probability financial instability
will occur within the following 3 months. It is possible to evaluate the probability of financial
instability P(C I  I t  I ) at different index values:

                                 A
        P(C I  I t  I ) 
                                A B                                                     (25)
where, I – lower interval limit, for which probability of financial instability is calculated; I –
upper interval limit; A – equals to number of cases when index value was in the interval [I ; I )
and within next 3 months crisis occurred; B equals to number of cases index value was in the
same interval [I ; I ) but crisis events didn’t occur.
        Table 3 presents results of empirical distribution of the probabilities of financial
instability in dependence from values of each index. Probability of financial instability decreases
fast and non-linear after indexes achieve relatively high values. In other words, if small amount

                                                                                                   61
of indicators signal, then probability of financial instability is low, although it sharply increases
after accumulation of alarm signals.


                                                                                                       Table 3
      Probabilities of the financial instability in dependence from the consolidated index values
                 I1                                         I2                               I3
Index value      Probabilities,%              Index value    Probabilities,%   Index value    Probabilities,%
     0                      0,00                 0-2              0,00           <0,05               0,00
     1                      2,17                 3-4              2,78          0,06-0,5             1,22
     2                      2,94                 5-8             12,50           0,5-0,7             25,00
     3                      12,50                 ≥9             66,67          0,7-1,53             40,00
     4                      20,00                                               1,53-1,6             50,00
     5                      40,00                                                 ≥1,6            100,00
     ≥6                     57,14


          After evaluation of the probabilities of financial instability depending on values of
consolidated indexes the problem of the selection of indicators with the best prognostic power
arises. For evaluation of prognostic power each of the indexes it is logical to evaluate deviation
of empirical probabilities of the financial instability Pt from actual probabilities (that is values of
dummy variable Rt which takes value 1 during three months before the crisis and 0 in all other
cases). Hence, for each of the indexes it is possible to calculate:
                      T
                 1
          Qk 
                 T
                      (P
                     t 1
                            t
                                k
                                    Rt ) 2
                                                                                              (26)
where, k = {1;2;3} is an index for which the prognostic power is calculated, and T equals to the
number of observations.
          Table 11 presents values of index Q for all three indexes and the best signaling indicator
(current account balance of the balance of payments) and variable equal to unconditional
probability of financial instability on during the entire examined period of time.




                                                                                                             62
                                                                                             Table 4
                                    Prognostic power of consolidated indexes of financial stability
                            Index                                     Prognostic power (Q)
                 Unconditional probability                                     0,134
   Current account balance of the balance of payments                          0,130
                              I1                                               0,093
                              I2                                               0,105
                              I3                                               0,074


       It is clear that the highest prognostic power has index I3 which takes in account
“operational capability” of each of indicator. Also, all consolidated indexes forecast financial
instability more precisely than the best of signaling indicators. However, this indicator is more
effective as a leading indicator of financial stability than being used for forecasting of
unconditional probability of financial instability.
       Thus, conducted analysis allows establishing a framework of signaling indicators of
financial instability which alarms in advance about developing problems on the Russian
financial market. On the basis of previous chapters it is possible to draw the following
conclusions.
       First, establishment of a framework of signaling indicators of financial instability is
conceptually possible. In other words, there are some series of the economic variables statistical
analysis allows to forecast occurrences of financial instability in case of the signals given by this
variables with the probability exceeding unconditional probability of the financial instability.
The application of different methodologies of the selection of working signaling indicators give
similar results and it confirms their effectiveness.
       Second, in spite of the relatively high working capability of the offered framework, most
of the indicators predict no more than half of the crisis episodes, increase of conditional
probability of financial instability in comparison with unconditional for some indicators does not
exceed 5-10%. That is why limitations of the offered methodology should be accounted and

                                                                                                  63
alternative estimation methods of the financial sector should be used, too.
       Third, my research showed that development of the framework of indicators of
financial instability for the particular country and not for the group of the countries allows
increases significantly the working capability of the indicators, which is especially important at
using offered methodology for decision-making by economic agents. Analysis of the
financial market of one country only accounts for its specific features and adopts accordingly
to threshold levels for the signaling indicators. Most working capabilities in the case of Russia
have shown the following indicators: current account balance for the balance of payments, real
interest rate, relation of money supply to gold and foreign currency reserves, real effective
exchange rate of ruble, and excessive money supply in real terms. Their use allows achieving an
increase of probability of financial instability in case of their signals in comparison with
unconditional probability on more than 40%.
       Finally, analysis allowed establishment of a consolidated index of financial stability I3
with help of which it is possible to obtain quantitative estimation of the occurrence of financial
instability. Probability of financial instability increases non-linearly in proportion to increase of
the number of signals given by indicators: if a small amount of indicators signal then a crisis
probability remains on the low level, though with the accumulation of the signals probability of
instability on the financial market sharply increases.
       In the last chapter an example of the application of the developed framework of
indicators for monitoring financial stability in Russian Federation 1999-2009 will be given.




                                                                                                  64
CHAPTER 4


Monitoring of the financial stability in Russia 2009 (II quarter)


        In this chapter I will approximate a system of signaling indicators of financial instability
on the financial markets of the developing countries andl examine the financial market of the
Russian Federation. The main obstacle to conducting the analysis is absence of the relevant
statistical data.
        Empirical models observed in the literature review offer a number of economic and
financial indicators for the sensitivity analysis of the analyzing country in relation to the crises
of financial system. Even though theory does not provide an unambiguous answer on what is an
early warning system, the performed research determines those variables which are supposed to
signal an approaching crisis.
        Offered analysis allows following up after long-term tendencies in the economy. Thus,
separate negative tendencies in a short-run may be compensated with favorable factors (for
example, in Russia that is a significant budget surplus caused by record oil prices). At the same
time a worsening of the market environment accumulated negative tendencies in economy might
cause financial instability.
        Main macro indicators of monitoring, which reflect overall situation in the financial sector
of the country are classified in the following order:
       Rate of economic growth: GDP growth ratio, dynamics of industrial output.
       Balance of payments: current account balance, gold and foreign currency reserves,
        foreign debt, terms of trade (export prices), export and import, real effective exchange
        rate, capital outflow.
       Interest rates: real exchange rate, difference between domestic and foreign interest rates,
        relation of credit rate to deposit rateм.
       Monetary indicators: customer price index, dynamics of the domestic credit in real terms,


                                                                                                 65
       monetary multiplier, growth ratio of deposits in real terms, relation of money supply to
       gold and foreign currency reserves, excessive money supply in real terms.
      Speculative pressure index
       If not pointed differently, in my work I will confine composition methodology to that
offered in the article of Kamisky, Lizondo and Reinhart (Kaminsky, Lizondo, Reinhart, 1998),
also I will follow closely the Manual of International Monetary Fund ( Compilation Guide on
Financial Soundness Indicators, 2006). Although during the process of selection of the relevant
indicators for the analysis, I took in account results of other similar researches.
       In most of the cases I used growth ratios of variables or analyze them in relation to GDP,
which allows ensuring data comparability and also solve the problem of non-stationary time
series. At the same time in a number of cases analysis uses variables expressed in levels because
in that particular way they have most prognostic ability. Table A5 of Appendix 5 shows the
description of changes of indicators, it also includes information about sources of information
and their periodicity.
       Dynamics for most of indicators is examined starting from 1999, when a majority of
negative consequences of the financial crisis in 1998 were already overcome. At the same time,
in a number of cases, mostly because of the absence of relevant statistical data, indicators may
be analyzed in a different time range.
       For the monitoring of financial stability in Russian Federation the methodology of
quantitative and qualitative analysis of the signaling indicators is used, within which threshold
levels of the indicators were established, and consolidated index of the financial stability was
calculated (which allows estimation of the probability of the financial instability during next
quarter).
       For a majority of variables there is no particular threshold level exceeding which will
allow speaking clearly about increase/decrease of the financial crisis probability. Analysis of
signaling indicators is strongly based on the expert evaluations of the current situation . That
is why making a final decision whether indicator signals about increase/decrease of the crisis
probability, is made through taking into consideration a particular current situation in the
economy of the country. Moreover, there are no particular values of indicators which would be
possible to call “normal” for the stable development of the financial system, because depending
on the particular state of economic situation the same value of indicator may either give reasons

                                                                                              66
to concern or be indicative of normal development.
       The consequence of analysis is following: first, the table summarizing monitoring results
on all variables is made. It is followed by the brief description of overall economic situation in
the country, and enclosure includes graphs of all indicators of the monitoring. Tables include
values of respective variables in the time line of few last quarters, allowing comparing changes.
Though they are used only for reference because besides absolute values of variables, while
dividing into groups of “good” and “bad” indicators I took in account seasonality of their
changes, previous dynamics, levels obtained and etc. Though, at the first sight they still allow to
draw some conclusions about current situation in economy of the country.

Russian Federation
        Results of the application of the quantity methodology are presented in the table below
which shows: the values of signaling indicators during the period of 2008-2009, threshold levels
and information of the signal. In Appendix 6 and 7 the dynamics of indicators is presented and
grap   hically demonstrated when indicators exceeded their threshold levels.
       In the first quarter of 2009 the financial system of the Russian Federation was faced
significant difficulties caused by capital outflow from the country, decline of the exchange rate
of ruble and a decrease in economic activity. Just like in the fourth quarter of 2008 in January, 5
out of 13 indicators signaled: GDP growth ratio in real terms, capital outflow, gold and foreign
currency reserves, export and speculative pressure index.
       In the beginning of 2009 it was clear that Russia would experience a full scale economic
crisis influenced not only by the financial sector of the economy but the real one too.
Government used funds accumulated over the years to soften the economic downturn. In spite of
some stabilization during spring, monitoring results show a high probability of new problems in
the financial system of Russian Federation within the next half a year. In all likelihood, these
problems might be caused by a further deterioration of world economy and domestic problems
related for example to growth of overdue payments of debts to the Russian banks.
       Due to an improvement of foreign economic environment in the second quarter of 2009,
the financial system of the Russian Federation was in better condition than in the beginning of
the year. In comparison to the first quarter, in April-June only 3 out 13 indicators gave signals:
GDP growth ratio in real terms, export and domestic credit in real terms. In my opinion these


                                                                                                  67
indicators demonstrate main threats which Russian economy may face in the short-run. In spite
of some pickup in the financial sector, the situation in the real sector is very complicated, and in
its turn, it will cause further growth of overdue payments on the loans issued to non-financial
sector by banks. Even in the case of recovery of the economy in the medium-term period
outrunning growth of import in comparison with exports most likely will cause a further decrease
of the current account position of the balance of payments, which will increase downward
pressure on the exchange rate of ruble. Moreover, in the conditions of recession increase of
budget expenditures in the end of the year will bring to the stage the problem of inflations.
        Thus, in spite of some stabilization in the Russian economy a worsening of the situation
in fall and winter 2009-2010 is very possible. Documented conditions of the framework of
indicators conform to the probability of the increase of scale of financial instability within the
next couple of quarters on the level of no more than 40%. This result means with high
probability of new problems in the financial sector of the Russian Federation within next half a
year.




                                                                                                 68
Conclusion


       In my research, the scientific evidences devoted to examination of the causes of financial
crises were examined. Principles of development of the processes individual for the Russian
financial market prior the episodes of financial instability were analyzed. Research was based on
the analysis of the framework of signaling indicators dynamics which allows a determination of
negative tendencies in the financial sector before the crisis.
       Presented research includes complex analysis of the problems related to the establishment
of the framework of signaling indicators and draws the following conclusions:
       1.      Conducted research created a list of working indicators which make sense to test
       on the conditions of the Russian financial market. This list includes the following
       variables: real exchange rate, GDP growth ratio, dynamics of the domestic credit, money
       supply, changes in the consumer prices, terms and trade among others. Examination of
       the scientific expertise within the defined field of research demonstrated limitations in the
       applicability of the offered methodology of forecasting financial instability. While
       established methodology permitted the determination of negative tendencies in the
       economy, they are not an absolutely accurate means of identification of the financial
       crisis. For a better picture of the situation in the financial sector of the country it is
       necessary to use expert evaluation and other methods of the analysis of the financial
       stability. At the same time offered methodology does not lose in quality or ability of
       forecasting to alternative approaches of the monitoring of financial stability.
       2.      Conducted analysis of the dynamics of signaling indicators defines a series of
       economic variables. The statistical analysis of these variables provides an opportunity to
       forecast occurrences of financial instability in the case of alarming signals with
       probability exceeding unconditional probability of the financial instability. Engagement
       of qualitative and non-parametric approaches of the indicators selection gave a similar
       result and all three methods complemented each other.
       3.      Establishment of the framework of signaling indicators of financial instability for


                                                                                                 69
a particular country instead of a group of countries, allows a significantly increase its
effectiveness, which is especially important for decision making by economic agents.
Analysis of the financial market of one country requires accounting for specific features
of the economy and adapting accordingly the threshold levels of the signaling indicators.
In the case of Russia, the most effective signaling indicators were current account balance
for the balance of payments, real interest rate, relation of money supply to gold and
foreign currency reserves, real effective exchange rate of the ruble and excessive money
supply in real terms. Their use allows with increasing probability of financial instability
in the case of the alarming signal in comparison with unconditional probability of
financial instability on the level of no more than 40%.
4.        The consolidated index of financial stability, through which it was possible to
obtain quantitative evaluation, was also established. With that, it turned out that the
probability of the occurrence of financial instability increased non-linearly in proportion
to the increasing number of the signals sent by signaling indicators.
5.        Offered methodology permits only a portion of information about developing
tendencies in the financial sector to be received but does not point out whether financial
crisis will happen or not. The offered analysis can be used to monitor medium-term
tendencies in the economy. Thus, negative tendencies in the short-run can be
compensated by favorable factors. At the same time, through a worsening of the
functional conditions, accumulated negative trends can cause financial instability.




                                                                                        70
Literature


1. Abiad A. Early Warning Systems: A Survey and a Regime-Switching Approach // IMF
Working Paper No.03/32. 2003.
2. Agenor P.-R., Bhandari J.S., Flood R.P. Speculative Attacks and Models of Balance-of-
Payments Crises. NBER Working Papers 3919, 1991.
3 Akerlof G., Romer P. Looting: The economic underworld of bankruptcy for profit // Brookings
Papers on Economic Activity. 1993.
4. Alexander W., Davis J., Ebrill L., Lindgren C.-J. Systemic Bank Restructuring and
Macroeconomic Policy. Washington, D.C.: IMF, 1997.
5. Altman E. Financial Ratios, Discriminant Analysis and the Prediction of the Corporate
Bankruptcy // Journal of Finance. 1968. Vol. 23 (September).
6. Arteta C., Eichengreen B. Banking Crises in Emerging Markets: Presumptions and Evidence //
Financial Policies in Emerging Markets (ed. by M. Blejer, M. Skreb). 2002.
7. Aziz J., F. Caramazza and R. Saldago. Currency crises: in search of common elements // IMF
working paper 00/67. March 2000.
8. Baig T., Goldfajn I. Financial Market Contagion in the Asian Crisis // IMF Staff Paper. 1999.
Vol. 46 (June).
9. Barseghyan L. Non-Performing Loans, Prospective Bailouts, and Japan’s Slowdown. The
Center for Japan-U.S. Business and Economic Studies Working Paper #317. Stern School of
Business, New York University. April 2004.
10. Bell J., Pain D. Leading indicator models of banking crises – a critical review // Financial
Stability Review. December 2000.
11. Berg A., Borenszstein E., G-M/ Milesi-Ferretti and C. Patillo. Anticipating balance of
payments crises – the role of early warning systems. IMF occasional paper No. 186. January
2000.




                                                                                             71
12. Berg A., Patillo C. Predicting Currency Crisis: The Indicators Approach and an Alternative //
Journal of International Money and Finance. 1999. Vol. 18.
13. Berg A., Patillo C. Are Currency Crises Predictable? A Test International Monetary Fund
Working Paper 98/154. November 1998.
14. Blanco H., Garber P. Recurrent Devaluation and Speculative Attacks on the Mexican Peso //
Journal of Political Economy, University of Chicago Press. Vol. 94(1). February 1986.
15. Brown C.O., Dinç S. The Politics of Bank Failures: Evidence from Emerging Markets. 2004.
16. Caballero R. and Krishnamurthy A. International and domestic collateral constraints in a
model of emerging market crises // Journal of Monetary Economics No. 48. 2001.
17. Caprio G., Honohan P. Restoring Banking Stability: Beyond Supervised Capital Requirement
// Journal of Economic Perspectives. 1999. Vol. 13. 4.
18. Caprio J., Klingebiel D. Bank insolvencies. Cross-country experience. World Bank Policy
Research Working paper. 1620. 1996a.
19. Caprio J., Klingebiel D. Bank insolvency: Bad luck, bad policy, or bad banking? Annual
World Bank Conference on Development Economics. 1996b.
20. Caprio J., Hunter W., Kaufmann G., Leipziger D. Preventing Bank Crisis. Lessons from
Recent Global Bank Failures. Washington, D.C.: The World Bank, 1998.
21. Caprio J. Safe and sound banking in developing countries: We’re not in Kansas more.
Prepared for the Brooking Conference ‘FDICIA: Bank Reform Five Years Later and Five Years
Ahead’ on December 19. 1996, 1997.
22. Caprio G., Martinez-Peria M.S. Avoiding Disaster: Policies to Reduce the Risk of Banking
Crises. Discussion Paper. 2000.
23. Caprio G., Summers L. Finance and Its Reform: Beyond Laissez- Faire’, Policy Research
Working Paper No. 1171. 1993.
24. Carramazza F., L. Ricci and R. Salgado. Trade and financial contagion in currency crises //
IMF working paper 00/55. March 2000.
25. Chari V., Jagannathan R. Banking panics, information, and rational expectations equilibrium
// Journal of Finance. 1988. 43.
26. Chen Y. Banking panics: The role of the first-come, first-served rule and information
externalities // Journal of Political Economy. 1999. 107.
27. Compilation Guide on Financial Soundness Indicators. IMF. July 2004.


                                                                                              72
28. Corsetti G., P. Pesenti and N. Roubini. Paper tigers? A model of the Asian crisis // NBER
Working Paper No. 6783. November 1998.
29. Cumby R.E., Wijnbergen S. van. Financial Policy and Speculative Runs with a Crawling
Peg: Argentina 1979–1981 // Journal of International Economics. Vol. 27. No.1/2. August 1989.
30. Davis E.P. Institutional Investors, Unstable Financial Markets and Monetary Policy // Risk
Management in Volatile Financial Markets (ed. By F. Bruni, D.E. Fair, R. O`Brien), 1996.
31. Demirguc-Kunt A., Detragiache E. Financial liberalization and financial fragility. Prepared
for the 1998 World Bank Annual Conference on Development Economics. 1998a.
32. Demirguc-Kunt A., Detragiache E. The determinants of banking crises in developing and
developed countries. IMF Staff Papers. 1998b. 45.
33. Demirguc-Kunt A., Detragiache E. Cross-Country Empirical Studies of Systemic Bank
Distress: A Survey. IMF Working Paper. May 2005.
34. Deutsche Bundesbank. The role of economic fundamentals in the emergence of currency
crises in emerging markets. Deutsche Bundesbank Monthly Report. April 1999.
35. Diamond D.W., Dybvig P.H. Bank Runs, Deposit Insurance, and Liquidity // Journal of
Political Economy. 1983. Vol. 91 (June).
36. Dornbusch R., Goldfajn I., Valdes R.O. Currency Crises and Collapses. Brookings Papers on
Economic Activity: Macroeconomics 2. Brookings Institution. 1995.
37. Dornbusch R. Expectations and Exchange Rate Dynamics // Journal of Political Economy.
1976. 84. December.
38. Dornbusch R. Collapsing Exchange Rate Regimes // Journal of Development Economics.
1987. Vol. 27. October.
39. Drees B., Pazarbasioglu C. The Nordic banking crisis. Pitfalls in financial liberalization?
IMF Occasional paper. 161. 1998.
40. Edwards S., Vegh C. Banks and macroeconomic disturbances under predetermined exchange
rate // Journal of Monetary Economics. 1997. 40.
41. Eichengreen B., Rose A. Staying afloat when the wind shifts: External factors and emerging-
market banking crises. NBER Working paper. 6370. 1998.
42. Eichengreen B., Rose A., Wyplosz C. Exchange market mayhem. The antecedents and
sftermath of speculative attacks // Economic Policy. 1995. October 1995.



                                                                                            73
43. Eichengreen B., Rose A., Wyplosz Ch. Contagious Currency Crises. NBER Working Paper
No. 5681. July 1996.
44. Enoch C., Green J. Banking System and Monetary Policy. Issues and Experience in the
Global Economy. Washington, D.C.: IMF, 1997.
45. Esquivel G. and F. Larrain. Explaining currency crisis. Harvard Institute for International
Development. June 1998.
46. Evans O., A. Leone, M. Gillard and R. Hilbers. Macroprudential indicators of financial
system soundness. IMF occasional paper No. 192, April 2000.
47. Fisher I. The Debt-Deflation Theory of Great Depression // Econometrica. 1933. Vol. 1
(October).
48. Flood R., Marion N. Perspectives of the Recent Currency Crisis Literature. NBER Working
Paper No. 6380 (Cambridge, Massachusetts, MIT Press). 1998.
49. Flood R., Garber P. Collapsing Exchange-Rate Regimes: Some Linear Examples // Journal of
International Economics. 1984. Vol. 17.
50. Flood R., Garber P. Linkages between Speculative Attack and Target Zone Models of
Exchange Rates // Quarterly Journal of Economics. 1991. Vol. 106.
51. Flood R.P., Marion N. Speculative Attacks: Fundamentals and Self- Fulfilling Prophecies.
NBER Working Paper 5789. 1996.
52. Flood R.P., Robert J. Hodrick. Real Aspects of Exchange Rate Regime Choice with
Collapsing Fixed Rates // Journal of International Economics. 1986. Vol. 21. November.
53. Flood R., Garber P., Kramer. Collapsing Exchange Rate Regimes: Another Linear Example
// Journal of International Economics. 1996. 41. No ¾. November.
54. Flood R., Marion N. A Model of the Joint Distribution of Banking and Exchange Rate Crises
// IMF Working Paper No. 01/213. 2001.
55. Flood R.P., Marion N. Perspectives on the Recent Currency Crisis Literature // International
Journal of Finance and Economics. 1999. Vol. 4. No. 1.
56. Frankel J.A., Rose A.K. Currency Crashes in Emerging Markets: An Empirical Treatment //
Journal of International Economics. 1996. Vol. 41 (November).
57. Frankel J.A., Rose A.K. Currency Crashes in Emerging Markets: Empirical Indicators.
NBER Working Paper No. 5437 (Cambridge, Massachusetts, MIT Press). 1996.



                                                                                             74
58. Fratzscher M. Why Are Currency Crises Contagious? A Comparison of the Latin American
Crisis of 1994-1995 and the Asian Crisis of 1997-1998 // Weltwirtschaftliches Archiv. 1998.
Vol. 134. 4.
59. Fratzscher M. What Causes Currency Crises: Sunspots, Contagion or Fundamentals?
European University Institute Department of Economics. EIU Working Papers 99/39. 1999.
60. Fratzscher M. On Currency Crises and Contagion // International Journal of Finance and
Economics. 2003. Vol. 8. No. 2.
61. Freixas X., Rochet J.-C. Microeconomics of Banking. Cambridge. MIT Press, 1997.
62. Gerlach S., Smets F. Contagious Speculative Attacks. CEPR Discussion Papers 1055.
C.E.P.R. Discussion Papers, 1994.
63. Glick R. and R. Moreno. Money and Credit, Competitiveness, and Currency Crises in Asia
and Latin America. Centre for Pacific Basin Monetary & Economic Studies working paper
PB99–01,
64. Goldberg L.S. Collapsing Exchange Rate Regimes: Shocks and Biases. NBER Working
Paper 2702. Cambridge, Massachusetts: National Bureau of Economic Research, 1998.
65. Goldberg L.S. Predicting Exchange Rate Crises: Mexico Revisited // Journal of international
Economics. 1994. 36.
66. Goldstein M., Kaminsky G.L., Reinhart C.M. Assessing Financial Vulnerability: An Early
Warning System for Emerging Markets. Institute for International Economics, 2000.
67. Goldstein M., Turner P. Banking crises in emerging economies: Origins and policy options.
BIS Economic Papers. 1996. 46.
68. Gonzalez-Hermosillo B., Pazarbasioglu C., Billings R. Determinants of banking system
fragility: A case study of Mexico. IMF Staff Papers. 1997. 44.
69. Guttentag J., Herring R. Credit Rationing and Financial Disorder // Journal of Finance. 1984.
Vol. 39 (December). Federal Reserve Bank of San Francisco. 1999.
70. Hardy D., Pazarbasioglu C. Leading indicators of banking crises: Was Asia different? IMF
Working paper. 98/91. 1998.
71. Hausmann R., Rojas-Suarez L. Banking Crises in Latin America. Washington, D.C.: IADB,
1996.
72. Hawkins J., Klau M. Measuring Potential Vulnerabilities in Emerging Market Economies.
BIS Working Paper 91. October 2000.


                                                                                              75
73. Hawkins J., Klau M. Early Warning Indicators for Emerging Economies. Paper prepared for
Irving Fisher Committee conference. 20–22 August 2002. Basel.
74. Hayashi F., Prescott E.C. The 1990s in Japan: A Lost Decade // Review of Economic
Dynamics, Academic Press for the Society for Economic Dynamics. Vol. 5(1). January 2002.
75. Heffernan S. An econometric model of bank failure // Economic and Financial Modelling.
Summer 1995.
76. Heffernan S. Modern Banking in Theory and Practice. John Wiley & Sons, 1996.
77. Herrera S. and C. Garcia. User`s Guide to an Early Warning System for Macroeconomic
Vulnerability in Latin American Countries. World Bank Working Paper 2233. November 1999.
78. Honohan P. Banking System Failures in Developing and Transition Countries: Diagnosis and
Prediction. BIS Working Paper . 39. 1997.
79. Hutchison M., McDill K. Are all banking crises alike? The Japanese experience in
international comparison. NBER Working paper. 7253. 1999.
80. International Monetary Fund. Chapter IV: Financial crises: characteristics and indicators of
vulnerability. World Economic Outlook. May 1998.
81. International Monetary Fund. International Capital Markets. September 1999.
82. International Monetary Fund. Early warning system models: the next steps forward // Global
Financial Stability Report. Chapter IV. March 2002.
83. Kamin S. and O. Babson. The contributions of domestic and external factors to Latin
American devaluation crisis: an early warning systems approach. US Federal Reserve System
International Finance Discussion Paper No. 645. September 1999.
84. Kaminsky G. Currency and banking crises: the early warnings of distress // IMF working
paper 99/178. December 1999.
85. Kaminsky G., Reinhart C. On Crises, Contagion, and Confusion // Journal of International
Economics. 2000. Vol. 51. Issue 1.
86. Kaminsky G., Reinhart C. The Twin Crises: The Causes of Banking and Balance-of-
Payments Problems // American Economic Review. 1999. Vol. 89 (June).
87. Kaminsky G., Lizondo S., Reinhart C. Leading Indicators of Currency Crises // IMF Staff
Papers. 1998. Vol. 45 (March).
88. Kaminsky G., Reinhart C. Financial Crises in Asia and Latin America: Then and Now //
AEA Papers and Proceedings. 1998. 98.


                                                                                             76
89. Kane E. J. The S&L Insurance Mess: How Did it Happen? Urban Institute Press. 1989.
90. Keeley M.C. Deposit Insurance, Risk and Market Power in Banking // American Economic
Review. 1990. Vol. 80 (December).
91. Kodres L., Pritsker M. A Rational Expectations Model of Financial Contagion. Finance and
Economics Discussion Series . 1998–48. 1998.
92. Kruger M., P. Osakwe and J. Page. Fundamentals, contagion and currency crises: an
empirical Analysis. Bank of Canada working paper No. 98–10, July 1998.
93. Krugman P. A Model of Balance-of-Payments Crises // Journal of Money, Credit and
Banking. 1979. Vol. 11 (August).
94. Krugman P. Target Zones and Exchange Rate Dynamics // Quarterly Journal of Economics.
1991. Vol. 106. August.
95. Krugman P., Rotemberg J. Target Zones with Limited Reserves. NBER Working Paper
3418. Cambridge, Massachusetts: National Bureau of Economic Research, 1990.
96. Krugman P. Balance Sheets, the Transfer Problem, and Financial Crises // International Tax
and Public Finance 6. 1999.
97. Lindgren C.-J., Garcia G., Saal M. Bank Soundness and Macroeconomic Policy.
Washington, D.C.: IMF, 1996.
98. Mehrez G., Kaufmann D. Transparency, Liberalization, and Financial Crises. 1999.
Milesi–Ferretti G. and A. Razin. Current Account Reversals and Currency Crises: Empirical
Regularities. IMF working paper No. 98/99. June 1998.
99. Minsky H. Theory of Systemic Fragility // In Altman E., Sametz A. Financial Crises:
Institutions and Markets in Fragile Environment. NY: Wiley, 1977.
100. Mishkin F.S. Understanding Financial Crises: A Developing Country Perspective. NBER
Working Paper 5600. 1996.
101. Muller C., R. Perrelli and M. Rocha. The role of corporate, legal and macroeconomic
balance sheet indicators in crisis detection and prevention // IMF working paper No. 02/59.
March 2002.
102. Obstfeld M. (1984). Balance-of-Payments Crises and Devaluation // Journal of Money,
Credit and Banking. Vol. 16. May.
103. Obstfeld M. The Logic of Currency Crises. NBER Working Paper 4640. Cambridge,
Massachusetts: National Bureau of Economic Research, 1994.


                                                                                           77
104. Obstfeld M. Models of Currency Crises with Self-Fulfilling Features. NBER Working
Paper 5285. Cambridge, Massachusetts: National Bureau of Economic Research, 1996.
105. Peresetsky .., Karminsky A., Golovan S. Probability of default models of Russian banks.
BOFIT Discussion Papers No 21. 2004.
106. Persaud A. Event risk indicator handbook. JP Morgan. London. January 1998.
107. Radelet S. and J. Sachs. The East Asian Financial Crisis: Diagnosis, Remedies, Prospects.
Brookings Papers on Economic Activity. 1998.
108. Rossi M. Financial Fragility and Economic Performance in Developing Economies: Do
Capital Controls, Prudential Regulation and Supervision Matter? // IMF working paper No.
99/66. May 1999.
109. Sachs J., Tornell A., Velasco A. Financial Crises in Emerging Markets: The Lessons from
1995. NBER Working Paper 5576. Cambridge, Massachusetts: National Bureau of Economic
Research, 1996.
110. Salant St., Henderson D. Market Anticipation of Government Policy and the Price of Gold
// Journal of Political Economy. 1978. 86.

111. Sargent T., Wallace N. Some unpleasant monetarist arithmetic // Federal Reserve Bank of
Minneapolis Quarterly Review. 1985. 9.

112. Sundararajan V., Balino T. Banking Crises: Cases and Issues. Washington, D.C.: IMF,
1991.

113. Sy A. Rating the Ratings Crises // IMF Working Paper No. 03/122. 2003.

114. Temzelides T. Evolution, coordination, and banking panics // Journal of Monetary
Economics. 1997. 40.

115. Tomczynska M. Early Indicators of Currency Crises. Review of some literature. Studies and
Analysis – Center for Social and Economic Research # 208 – Warsaw, 2000.

116. Tornell A. Common fundamentals in the Tequila and Asian crises // NBER Working Paper
No. 7139. 1999.




                                                                                           78

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:95
posted:5/19/2012
language:
pages:78
fanzhongqing fanzhongqing http://
About